Search

Your search keyword '"Ghavamzadeh A"' showing total 3,103 results
Start Over Author "Ghavamzadeh A"
3,103 results on '"Ghavamzadeh A"'

1. Q-learning for Quantile MDPs: A Decomposition, Performance, and Convergence Analysis

Author

Hau, Jia Lin, Delage, Erick, Derman, Esther, Ghavamzadeh, Mohammad, and Petrik, Marek

Subjects
Computer Science - Machine Learning
Abstract

In Markov decision processes (MDPs), quantile risk measures such as Value-at-Risk are a standard metric for modeling RL agents' preferences for certain outcomes. This paper proposes a new Q-learning algorithm for quantile optimization in MDPs with strong convergence and performance guarantees. The algorithm leverages a new, simple dynamic program (DP) decomposition for quantile MDPs. Compared with prior work, our DP decomposition requires neither known transition probabilities nor solving complex saddle point equations and serves as a suitable foundation for other model-free RL algorithms. Our numerical results in tabular domains show that our Q-learning algorithm converges to its DP variant and outperforms earlier algorithms.

Published
2024

2. Confidence-aware Reward Optimization for Fine-tuning Text-to-Image Models

Author

Kim, Kyuyoung, Jeong, Jongheon, An, Minyong, Ghavamzadeh, Mohammad, Dvijotham, Krishnamurthy, Shin, Jinwoo, and Lee, Kimin

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Fine-tuning text-to-image models with reward functions trained on human feedback data has proven effective for aligning model behavior with human intent. However, excessive optimization with such reward models, which serve as mere proxy objectives, can compromise the performance of fine-tuned models, a phenomenon known as reward overoptimization. To investigate this issue in depth, we introduce the Text-Image Alignment Assessment (TIA2) benchmark, which comprises a diverse collection of text prompts, images, and human annotations. Our evaluation of several state-of-the-art reward models on this benchmark reveals their frequent misalignment with human assessment. We empirically demonstrate that overoptimization occurs notably when a poorly aligned reward model is used as the fine-tuning objective. To address this, we propose TextNorm, a simple method that enhances alignment based on a measure of reward model confidence estimated across a set of semantically contrastive text prompts. We demonstrate that incorporating the confidence-calibrated rewards in fine-tuning effectively reduces overoptimization, resulting in twice as many wins in human evaluation for text-image alignment compared against the baseline reward models., Comment: ICLR 2024

Published
2024

3. Bridging Distributionally Robust Learning and Offline RL: An Approach to Mitigate Distribution Shift and Partial Data Coverage

Author

Panaganti, Kishan, Xu, Zaiyan, Kalathil, Dileep, and Ghavamzadeh, Mohammad

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

The goal of an offline reinforcement learning (RL) algorithm is to learn optimal polices using historical (offline) data, without access to the environment for online exploration. One of the main challenges in offline RL is the distribution shift which refers to the difference between the state-action visitation distribution of the data generating policy and the learning policy. Many recent works have used the idea of pessimism for developing offline RL algorithms and characterizing their sample complexity under a relatively weak assumption of single policy concentrability. Different from the offline RL literature, the area of distributionally robust learning (DRL) offers a principled framework that uses a minimax formulation to tackle model mismatch between training and testing environments. In this work, we aim to bridge these two areas by showing that the DRL approach can be used to tackle the distributional shift problem in offline RL. In particular, we propose two offline RL algorithms using the DRL framework, for the tabular and linear function approximation settings, and characterize their sample complexity under the single policy concentrability assumption. We also demonstrate the superior performance our proposed algorithm through simulation experiments., Comment: 33 pages, preprint

Published
2023

4. Preference Elicitation with Soft Attributes in Interactive Recommendation

Author

Biyik, Erdem, Yao, Fan, Chow, Yinlam, Haig, Alex, Hsu, Chih-wei, Ghavamzadeh, Mohammad, and Boutilier, Craig

Subjects
Computer Science - Information Retrieval, Computer Science - Artificial Intelligence
Abstract

Preference elicitation plays a central role in interactive recommender systems. Most preference elicitation approaches use either item queries that ask users to select preferred items from a slate, or attribute queries that ask them to express their preferences for item characteristics. Unfortunately, users often wish to describe their preferences using soft attributes for which no ground-truth semantics is given. Leveraging concept activation vectors for soft attribute semantics, we develop novel preference elicitation methods that can accommodate soft attributes and bring together both item and attribute-based preference elicitation. Our techniques query users using both items and soft attributes to update the recommender system's belief about their preferences to improve recommendation quality. We demonstrate the effectiveness of our methods vis-a-vis competing approaches on both synthetic and real-world datasets.

Published
2023

5. Factual and Personalized Recommendations using Language Models and Reinforcement Learning

Author

Jeong, Jihwan, Chow, Yinlam, Tennenholtz, Guy, Hsu, Chih-Wei, Tulepbergenov, Azamat, Ghavamzadeh, Mohammad, and Boutilier, Craig

Subjects
Computer Science - Artificial Intelligence
Abstract

Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users.

Published
2023

6. Bayesian Regret Minimization in Offline Bandits

Author

Petrik, Marek, Tennenholtz, Guy, and Ghavamzadeh, Mohammad

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

We study how to make decisions that minimize Bayesian regret in offline linear bandits. Prior work suggests that one must take actions with maximum lower confidence bound (LCB) on their reward. We argue that the reliance on LCB is inherently flawed in this setting and propose a new algorithm that directly minimizes upper bounds on the Bayesian regret using efficient conic optimization solvers. Our bounds build heavily on new connections to monetary risk measures. Proving a matching lower bound, we show that our upper bounds are tight, and by minimizing them we are guaranteed to outperform the LCB approach. Our numerical results on synthetic domains confirm that our approach is superior to LCB.

Published
2023

7. Pericardial Disease in HSCT

Author

Ghavamzadeh, Ardeshir, Emami, Amir Hossein, Roudini, Kamran, Rezaei Kalantari, Kiara, Mohseni, Mina, Jafari Fesharaki, Mehrdad, Alizadehasl, Azin, editor, Ghavamzadeh, Ardeshir, editor, Emami, Amir Hossein, editor, Janbabaei, Ghasem, editor, and Khoda-Amorzideh, Davood, editor

Published
2024
Full Text
View/download PDF

8. HSCT at a Glance

Author

Ghavamzadeh, Ardeshir, Barkhordar, Maryam, Alizadehasl, Azin, editor, Ghavamzadeh, Ardeshir, editor, Emami, Amir Hossein, editor, Janbabaei, Ghasem, editor, and Khoda-Amorzideh, Davood, editor

Published
2024
Full Text
View/download PDF

9. DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models

Author

Fan, Ying, Watkins, Olivia, Du, Yuqing, Liu, Hao, Ryu, Moonkyung, Boutilier, Craig, Abbeel, Pieter, Ghavamzadeh, Mohammad, Lee, Kangwook, and Lee, Kimin

Subjects
Computer Science - Machine Learning, Computer Science - Computer Vision and Pattern Recognition
Abstract

Learning from human feedback has been shown to improve text-to-image models. These techniques first learn a reward function that captures what humans care about in the task and then improve the models based on the learned reward function. Even though relatively simple approaches (e.g., rejection sampling based on reward scores) have been investigated, fine-tuning text-to-image models with the reward function remains challenging. In this work, we propose using online reinforcement learning (RL) to fine-tune text-to-image models. We focus on diffusion models, defining the fine-tuning task as an RL problem, and updating the pre-trained text-to-image diffusion models using policy gradient to maximize the feedback-trained reward. Our approach, coined DPOK, integrates policy optimization with KL regularization. We conduct an analysis of KL regularization for both RL fine-tuning and supervised fine-tuning. In our experiments, we show that DPOK is generally superior to supervised fine-tuning with respect to both image-text alignment and image quality. Our code is available at https://github.com/google-research/google-research/tree/master/dpok., Comment: NeurIPS 2023

Published
2023

10. Private and Communication-Efficient Algorithms for Entropy Estimation

Author

Bravo-Hermsdorff, Gecia, Busa-Fekete, Róbert, Ghavamzadeh, Mohammad, Medina, Andres Muñoz, and Syed, Umar

Subjects
Computer Science - Machine Learning, Computer Science - Cryptography and Security, Computer Science - Information Theory, Mathematics - Statistics Theory
Abstract

Modern statistical estimation is often performed in a distributed setting where each sample belongs to a single user who shares their data with a central server. Users are typically concerned with preserving the privacy of their samples, and also with minimizing the amount of data they must transmit to the server. We give improved private and communication-efficient algorithms for estimating several popular measures of the entropy of a distribution. All of our algorithms have constant communication cost and satisfy local differential privacy. For a joint distribution over many variables whose conditional independence is given by a tree, we describe algorithms for estimating Shannon entropy that require a number of samples that is linear in the number of variables, compared to the quadratic sample complexity of prior work. We also describe an algorithm for estimating Gini entropy whose sample complexity has no dependence on the support size of the distribution and can be implemented using a single round of concurrent communication between the users and the server. In contrast, the previously best-known algorithm has high communication cost and requires the server to facilitate interaction between the users. Finally, we describe an algorithm for estimating collision entropy that generalizes the best known algorithm to the private and communication-efficient setting., Comment: Originally published at the 36th Conference on Neural Information Processing Systems (NeurIPS 2022). This version corrects some errors in the original version

Published
2023

11. On Dynamic Programming Decompositions of Static Risk Measures in Markov Decision Processes

Author

Hau, Jia Lin, Delage, Erick, Ghavamzadeh, Mohammad, and Petrik, Marek

Subjects
Mathematics - Optimization and Control, Computer Science - Artificial Intelligence
Abstract

Optimizing static risk-averse objectives in Markov decision processes is difficult because they do not admit standard dynamic programming equations common in Reinforcement Learning (RL) algorithms. Dynamic programming decompositions that augment the state space with discrete risk levels have recently gained popularity in the RL community. Prior work has shown that these decompositions are optimal when the risk level is discretized sufficiently. However, we show that these popular decompositions for Conditional-Value-at-Risk (CVaR) and Entropic-Value-at-Risk (EVaR) are inherently suboptimal regardless of the discretization level. In particular, we show that a saddle point property assumed to hold in prior literature may be violated. However, a decomposition does hold for Value-at-Risk and our proof demonstrates how this risk measure differs from CVaR and EVaR. Our findings are significant because risk-averse algorithms are used in high-stake environments, making their correctness much more critical.

Published
2023

12. A Review of Deep Learning for Video Captioning

Author

Abdar, Moloud, Kollati, Meenakshi, Kuraparthi, Swaraja, Pourpanah, Farhad, McDuff, Daniel, Ghavamzadeh, Mohammad, Yan, Shuicheng, Mohamed, Abduallah, Khosravi, Abbas, Cambria, Erik, and Porikli, Fatih

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Video captioning (VC) is a fast-moving, cross-disciplinary area of research that bridges work in the fields of computer vision, natural language processing (NLP), linguistics, and human-computer interaction. In essence, VC involves understanding a video and describing it with language. Captioning is used in a host of applications from creating more accessible interfaces (e.g., low-vision navigation) to video question answering (V-QA), video retrieval and content generation. This survey covers deep learning-based VC, including but, not limited to, attention-based architectures, graph networks, reinforcement learning, adversarial networks, dense video captioning (DVC), and more. We discuss the datasets and evaluation metrics used in the field, and limitations, applications, challenges, and future directions for VC., Comment: 42 pages, 10 figures

Published
2023

13. The Effect of Conjugated Linoleic Acid Supplementation on Lipid-Related Cardiovascular Biomarkers in Obese Adults

Author

Fatemeh Esmaeili Shahmirzadi, Saeid Ghavamzadeh, and Arash Rashidi

Subjects
linoleic acid, conjugated linoleic acid: cardiovascular risk, obesity, Agriculture, Nutrition. Foods and food supply, TX341-641
Abstract

Studies have shown incompatible findings regarding the effects of conjugated linoleic acid (CLA) supplementation on cardiovascular diseases (CVDs) risk factors. The aim of this study was to evaluate the effect of daily CLA supplementation on serum insulin and lipid- related CV biomarkers in obese adults. Methods: This randomized double-blind clinical trial was conducted on 54 adults categorized as class I obesity. The participants were randomly assigned into two groups (n=27) receiving a total of 3,000 mg/d of a 50:50 mixture of CLA isomers for three months in intervention group (IG) and 500 mg/d paraffin in placebo group (PG). Moreover, fasting serum levels of insulin, lipid profile, non-HDL-Cholesterol (non-HDL-C), atherogenic index of plasma (AIP), total triglyceride (TG)/HDL-C, and cholesterol/HDL-C ratio were measured. The main statistical analysis method was independent t-test for changes. Results: Changes between the groups showed a significant decrease in total cholesterol (P=0.03), LDL-C (P=0.04), and non-HDL-C (P=0.03), and also a significant increase in AIP (P=0.04) in IG compared to the PG. A remarkable decrease was found in HDL-C and cholesterol/HDL-C ratio. In addition, a remarkable increase was observed in TG in this context. Serum insulin, VLDL-C, and LDL-C/HDL-C ratio showed no significant changes during the intervention period. The use of CLA supplementation could help reduce some adverse fractions of serum lipid profile, particularly TC, non-HDL-C and LDL-C. Conclusions: Regarding the augmenting effects of CLA intake on AIP as a strong predictive marker for CVDs, it is difficult to confirm the beneficial effects of CLA supplementation in preventing CVDs.

Published
2024

14. Aligning Text-to-Image Models using Human Feedback

Author

Lee, Kimin, Liu, Hao, Ryu, Moonkyung, Watkins, Olivia, Du, Yuqing, Boutilier, Craig, Abbeel, Pieter, Ghavamzadeh, Mohammad, and Gu, Shixiang Shane

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computer Vision and Pattern Recognition
Abstract

Deep generative models have shown impressive results in text-to-image synthesis. However, current text-to-image models often generate images that are inadequately aligned with text prompts. We propose a fine-tuning method for aligning such models using human feedback, comprising three stages. First, we collect human feedback assessing model output alignment from a set of diverse text prompts. We then use the human-labeled image-text dataset to train a reward function that predicts human feedback. Lastly, the text-to-image model is fine-tuned by maximizing reward-weighted likelihood to improve image-text alignment. Our method generates objects with specified colors, counts and backgrounds more accurately than the pre-trained model. We also analyze several design choices and find that careful investigations on such design choices are important in balancing the alignment-fidelity tradeoffs. Our results demonstrate the potential for learning from human feedback to significantly improve text-to-image models.

Published
2023

15. Offline Reinforcement Learning for Mixture-of-Expert Dialogue Management

Author

Gupta, Dhawal, Chow, Yinlam, Tulepbergenov, Aza, Ghavamzadeh, Mohammad, and Boutilier, Craig

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computation and Language
Abstract

Reinforcement learning (RL) has shown great promise for developing dialogue management (DM) agents that are non-myopic, conduct rich conversations, and maximize overall user satisfaction. Despite recent developments in RL and language models (LMs), using RL to power conversational chatbots remains challenging, in part because RL requires online exploration to learn effectively, whereas collecting novel human-bot interactions can be expensive and unsafe. This issue is exacerbated by the combinatorial action spaces facing these algorithms, as most LM agents generate responses at the word level. We develop a variety of RL algorithms, specialized to dialogue planning, that leverage recent Mixture-of-Expert Language Models (MoE-LMs) -- models that capture diverse semantics, generate utterances reflecting different intents, and are amenable for multi-turn DM. By exploiting MoE-LM structure, our methods significantly reduce the size of the action space and improve the efficacy of RL-based DM. We evaluate our methods in open-domain dialogue to demonstrate their effectiveness w.r.t.\ the diversity of intent in generated utterances and overall DM performance., Comment: Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023)

Published
2023

16. Multi-Task Off-Policy Learning from Bandit Feedback

Author

Hong, Joey, Kveton, Branislav, Katariya, Sumeet, Zaheer, Manzil, and Ghavamzadeh, Mohammad

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Many practical applications, such as recommender systems and learning to rank, involve solving multiple similar tasks. One example is learning of recommendation policies for users with similar movie preferences, where the users may still rank the individual movies slightly differently. Such tasks can be organized in a hierarchy, where similar tasks are related through a shared structure. In this work, we formulate this problem as a contextual off-policy optimization in a hierarchical graphical model from logged bandit feedback. To solve the problem, we propose a hierarchical off-policy optimization algorithm (HierOPO), which estimates the parameters of the hierarchical model and then acts pessimistically with respect to them. We instantiate HierOPO in linear Gaussian models, for which we also provide an efficient implementation and analysis. We prove per-task bounds on the suboptimality of the learned policies, which show a clear improvement over not using the hierarchical model. We also evaluate the policies empirically. Our theoretical and empirical results show a clear advantage of using the hierarchy over solving each task independently., Comment: 14 pages, 3 figures

Published
2022

17. ARMS-PCR Versus AS-PCR to evaluate JAK2V617F mutation in patients with non-CML myeloproliferative neoplasms

Author

Nadali F, Ferdowsi Sh, Karimzadeh P, Chahardouli B, Einollahi N, Mousavi A, Bahar B, Dargahi H, Toogeh GhR, Alimoghaddam K, Ghavamzadeh A, and Ghaffari SH

Subjects
Mutation, myeloproliferative, neoplasm, PCR, Medicine (General), R5-920
Abstract

"n Normal 0 false false false EN-US X-NONE AR-SA MicrosoftInternetExplorer4 /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-qformat:yes; mso-style-parent:""; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin:0cm; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:11.0pt; font-family:"Calibri","sans-serif"; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman"; mso-fareast-theme-font:minor-fareast; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin; mso-bidi-font-family:Arial; mso-bidi-theme-font:minor-bidi;} Background: JAK2 is a nonreceptor tyrosine kinase that plays a major role in myeloid disorders. This mutation is characterized by a G to T transverse at nucleotide 1849 in exon 12 of the JAK2 gene, located on the chromosome 9p, leading to a substitution of valine to phenylalanine at amino acid position 617 in the JAK2 protein. In this study we compared the amplification refractory mutation (ARMS) assay and allele- specific (AS-PCR) to evaluate JAK2V617F mutation patients with non-CML myeloproliferative neoplasms (MPNS)."n"nMethods: In this experimental study we evaluated JAK2 mutation in 58 patients with a known or suspected diagnosis of a myeloproliferative neoplasm by simple randomized sampling. The mutation was detected by ARMS-PCR and AS-PCR in patients. In order to verify the methods, amplified products from some patients were sequenced."n"nResults: The JAK2 V617F mutation was detected in 86.6%(26/30) of patients with polycythemia vera and 61.5%(8/13) of patients with idiopathic myelofibrosis by ARMS-PCR and AS-PCR. 46.6%(7.15) of essential thrombocythemia patients were positive using ARMS- PCR method while 53%(8.15) of then were positive when AS- PCR were used. The mutation was confirmed by sequencing."n"nConclusions: The incidence of JAK2 mutation using above PCR methods is similar to previous studies. The different results may depend on the molecular technique used.

Published
2010

18. Development of a quantitative Real-Time PCR for micrometastasis detection using CEA in peripheral blood and bone marrow specimens of gastric cancer patients

Author

Dardaei Alghalandis L, Shahsavani R, Ghavamzadeh A, Behmanesh M, and Aslankoohi E

Subjects
Carcino embryonic antigen, gastric, adenocarcinoma, metastasis, PCR, Medicine (General), R5-920
Abstract

"n Normal 0 false false false EN-US X-NONE AR-SA MicrosoftInternetExplorer4 /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-qformat:yes; mso-style-parent:""; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-para-margin:0in; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:11.0pt; font-family:"Calibri","sans-serif"; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman"; mso-fareast-theme-font:minor-fareast; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin; mso-bidi-font-family:Arial; mso-bidi-theme-font:minor-bidi;} Background: Gastric adenocarsinoma is the first leading fatal malignancy in Iran. Despite advances in novel therapeutics approaches for gastric cancer (GC) patient, tumor dissemination via blood stream to distant organ is still the major cause of death. Therefore, there is urgent need to establish sensitive methods for early detection of disseminated tumor cells in peripheral blood (PB) and bone marrow (BM) specimens of gastric cancer patients. "n"nMethods: In the present study, we use Carcinoma Embryonic Antigen (CEA) as a tumor marker and Glyceraldehyde 3-Phosphate Dehydrogenase (GAPDH) as an internal control to detection and quantification of disseminated tumor cells in PB and BM specimens of affected individuals. Total RNA was extracted from AGS (gastric cancer) cell line and CEA and GAPDH fragments were generated by reverse transcription. The amplified fragments were cloned into pTZ57R/T vector separately. Double cloning of these genes has done into one pTZ57R/T vector. Serial dilution of this recombinant plasmid is used to construct standard curve, each containing a known amount of input copy number. Total RNA was extracted from BP and BM specimens of 35 GC patients. cDNA of the specimens were synthesized by reverse transcription and subjected to Quantitative Real-Time PCR (QRT-PCR)."n"nResults: We developed a highly sensitive and specific quantitative PCR for CEA and GAPDH using Real-Time PCR based on TaqMan technology. CEA mRNA was detected in 23% of PB and 20% of BM specimens. There was no CEA mRNA detecting in control group."n"nConclusions: The QRT-PCR for CEA can be a useful technique for detection of micrometastases in the PB and BM specimens of gastric cancer patients."n

Published
2009

19. Detection of bladder transitional cell carcinoma: urinary hTERT assay versus urine cytology

Author

Yahyazadeh SR, Mehraban D, Ghaffari SH, Alimoghadam K, Ghavamzadeh A, Naderi Gh, Kazemeyni SM, and Rasteh M

Subjects
Bladder carcinoma, transitional cell carcinoma, telomerase, urine cytology, RT-PCR, Medicine (General), R5-920
Abstract

"nBackground: Transitional Cell Carcinoma (TCC) of bladder is the second most common urogenital malignancy and because of its high rate of recurrence (two third of tumors recur) vigilant surveillance is necessary. There have been a lot of efforts to find a proper biomarker for detecting urothelial cancers because available methods are expensive and invasive (like cystoscopy) or have a low degree of sensitivity (like urine cytology). Urothelial malignancies, like other cancers tend to express a large amount of telomerase. The aim of this study was to evaluate the possible application of voided urine human telomerase reverse transcriptase (hTERT) mRNA assay in detecting low-grade bladder carcinoma in comparison with urine cytology. "nMethods: Voided urine samples were collected from 49 patients who were supposed to go under operation. Samples were examined by both Quantitative Real-time RT-PCR (for measuring hTERT mRNA level) and cytology; the results were then compared to the final pathologic studies. "nResults: Regardless of clinical stage and or pathological grade of tumor, sensitivity of telomerase test and urine cytology was 74% and 16% respectively. There was a strong correlation between results of urine cytology and stage and/or grade of tumor; however, sensitivity of telomerase test was acceptable regardless of stage and or grade of tumor. There was a statistically significant difference between sensitivity of urine cytology and telomerase test (p

Published
2009

20. Operator Splitting Value Iteration

Author

Rakhsha, Amin, Wang, Andrew, Ghavamzadeh, Mohammad, and Farahmand, Amir-massoud

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Electrical Engineering and Systems Science - Systems and Control, Mathematics - Optimization and Control, Statistics - Machine Learning
Abstract

We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error., Comment: Accepted to NeurIPS2022

Published
2022

21. RASR: Risk-Averse Soft-Robust MDPs with EVaR and Entropic Risk

Author

Hau, Jia Lin, Petrik, Marek, Ghavamzadeh, Mohammad, and Russel, Reazul

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Prior work on safe Reinforcement Learning (RL) has studied risk-aversion to randomness in dynamics (aleatory) and to model uncertainty (epistemic) in isolation. We propose and analyze a new framework to jointly model the risk associated with epistemic and aleatory uncertainties in finite-horizon and discounted infinite-horizon MDPs. We call this framework that combines Risk-Averse and Soft-Robust methods RASR. We show that when the risk-aversion is defined using either EVaR or the entropic risk, the optimal policy in RASR can be computed efficiently using a new dynamic program formulation with a time-dependent risk level. As a result, the optimal risk-averse policies are deterministic but time-dependent, even in the infinite-horizon discounted setting. We also show that particular RASR objectives reduce to risk-averse RL with mean posterior transition probabilities. Our empirical results show that our new algorithms consistently mitigate uncertainty as measured by EVaR and other standard risk measures.

Published
2022

22. Bone Densitometric changes after bone marrow transplantation in 63 patients with leukemia and lymphoma

Author

Esfahani A, Iravani M, Khoshnyat M, Ghoreishi Z, Shamshiri A R, Moghadam Z, Jahani M, and Ghavamzadeh A

Subjects
Medicine (General), R5-920
Abstract

Background: Bone marrow transplantation (BMT) is the treatment of choice for many patients with malignant and nonmalignant diseases. Long-term complications such as osteoporosis should be considered, because it is directly associated with the morbidity and mortality. The purpose of this study is to assess the bone mineral density after allogenic or autologous bone marrow transplantation in patients with leukemia or lymphoma.Methods: We prospectively investigated 63 patients undergoing BMT for acute and chronic leukemia and lymphoma. At the end of the study, a total of 28 patients were assessed. Bone mineral density (BMD) was measured prior BMT, and 6 and 12 months after BMT. Osteocalcin, bone alkaline phosphatase and C-terminal telopeptides of type 1 collagen (ICTP) were assessed. Serum concentration of calcium, phosphorous, vitamin D, PTH and sex hormones (FSH, LH, testosterone and estradiol) were also measured.Results: There was a significant decrease in the bone mineral density of the femoral neck six months after BMT (p

Published
2007

23. Prognostic Significance of Circulating and Disseminated Tumor Cells in Breast Cancer Patients before and after Adjuvant Chemotherapy

Author

Parisa Ghaffari, Meysam Yousefi, Mozaffar Aznab, Negar Khazan, Marjan Yaghmaie, Davood Bashash, Mohammad Vaezi, Ardashir Ghavamzadeh, and Seyed H Ghaffari

Subjects
breast cancer, circulating tumor cells, disseminated tumor cells, real-time polymerase chain reaction, Medicine, Science
Abstract

Objective: Despite the advances in treatment, breast cancer (BC) remains a major cause of death in women. Thisstudy aims to evaluate the prognostic significance of detecting circulating tumor cells (CTCs) and disseminated tumorcells (DTCs) in paired peripheral blood (PB) and bone marrow (BM) samples obtained both before and after adjuvantchemotherapy from patients with operable BC.Materials and Methods: In this experimental study, from 160 patients with primary BC, we collected 160 PB and BM samplesbefore and we could be able to collect PB and BM samples from 100 of them after adjuvant chemotherapy. The expressionlevel of cytokeratin 19 (CK19), carcinoembryonic antigen (CEA), mammaglobin 1 (MGB1), mucin 2 (MUC2) and trefoil factor1 (TFF1) mRNAs in the PB/BM samples were analyzed by quantitative real-time polymerase chain reaction (PCR).Results: Multivariate Cox regression analyses indicated that the detection of CK19 mRNA-positive CTCs/DTCs eitherbefore or after adjuvant chemotherapy was an independent factor for prognosis associated with decreased diseasefreesurvival (DFS). Patients with tumor cells detected in both PB and BM and patients with persistent detection oftumor cells before and after chemotherapy had worse outcomes compared to those with tumor cells detected in one orneither of the compartments.Conclusion: This study suggests that the detection of CK19 mRNA-positive CTCs/DTCs either before or after adjuvantchemotherapy could be an independent predictor of DFS in operable BC patients.

Published
2024
Full Text
View/download PDF

24. Clinical, Biological and Pathological Characteristics of Breast Cancer Patients at the Taleghani University Hospital in Kermanshah, Iran

Author

Shahriari Ahmadi A., Ghavamzadeh A., Amiri N., Farnia V., Samadzadeh S., and Malekniazi A.

Subjects
Biological Clinical and Pathological Characteristics, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Introduction: Breast cancer is the most common of all malignant neoplasms in women worldwide. This study aims to demonstrate certain biological, clinical and pathological characteristics of patients treated at the university hospital oncology unit. Methods: A descriptive study was conducted during a period of 2 years, from October 2003 through September of 2005 in Kermanshah, Iran. 555 patients were selected to participate, representing all the cases diagnosed and treated for breast cancer. Data was gathered according to questionnaires and pa-tients’ records. Results: The mean age at which breast cancer was first diagnosed was 46.5±11.6 year of age with 89% of tumors being infiltrating intraductal carcinoma. The majority of the patient population had tumors stage II and grade II. Mean tumor size was 2.14±0.57 centimeters. 58% of the tumors were localized to the upper outer quadrant of the affected breast and 89% of the patients received modified radical mastectomies with almost a 92% two year survival. Conclusion: Highest prevalence of breast cancer was recorded in the 40-49 (mean 46) years of age group which compares favorably with studies done under similar circumstances. Tumor size, grade, stage, tumor marker analysis, metastasis and other disease characteristics portray patient population tendencies for breast cancer patients in Kermanshah, Iran.

Published
2005

25. Frequency of BCR-ABL Fusion Transcript in Iranian Patients with Chronic Myeloid Leukemia

Author

Yaghmaie M., Ghaffari S.H., Alimoghaddam K., Ghavamzadeh A., Mousavi S.A., Irvani M., Bahar B., Bibordi E., and Jahani M

Subjects
BCR-ABL, CML, Multiplex RT-PC, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Introduction: Reverse transcriptase-polymerase chain reaction (RT-PCR) assay is a useful tool for the detection of fusion transcript resulting from specific chromosomal translocation of the leukemia cells. A specific chromosomal abnormality, the Philadelphia chromosome (Ph), is present in 90% to 95% of CML patients.The aberration results from a reciprocal translocation between chromosome 9 and 22, creating a BCR-ABL fusion gene.There are two major forms of the BCR/ABL fusion gene, involving ABL exon 2, but including different exons of BCR gene. The transcripts b2a2 or b3a2 code for a p210 protein. Another fusion gene leads to the expression of an e1a2 transcript, which codes for a p190 pro-tein. Another, less common fusion genes are b3a3 or b2a3 (p203) and e19a2 (p230). The incidence of one or other rearrangement in chronic myeloid leukemia (CML) patients varies in different reported se-ries. In general, fusion transcripts are determined individually, a process which is labor intensive in or-der to detect all major fusion transcripts. Methods: This study was designed to determine the frequency of different fusion genes in 75 iranian patients with CML. peripheral blood samples were analyzed by multiplex reverse transcriptase poly-merase chain reaction (RT-PCR) from adult patients to detect all types of BCR-ABL transcripts of the t (9:22) and found that all cases were positive for some type of BCR/ABL rearrangement. Results: Most of our patients showed b3a2 fusion gene (62%), while the remaining showed one of the transcripts of b2a2, b3a3, b2a3, e1a2 or coexpression of b3a2 and b2a2. The rate of coexpression of the b3a2 and b2a2 was 5%. Conclusion: In contrast to the other reports, we did not see any coexpression of p210/p190. This may reflect either the sensitivity of the detection techniques used or the possibility of genetic differences be-tween the populations studied. Coexpression may be due to alternative splicing or to phenotypic varia-tion, with clinical course different from classical CML.

Published
2005

26. Early Hepatic Complication in First Year after Bone Marrow Transplantation in Major Beta Thalassemic Patients

Author

Iravani M, Arshy M, Toutounchi M, Nedaeifard L, and Ghavamzadeh A

Subjects
Hepatic complications, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Introduction: Bone marrow transplantation is a good therapeutic modality for beta thalassemia. Liver complications are one of the major causes of morbidity and mortality following BMT. Determination of the factors of liver injury leads to earlier diagnosis after BMT and improves prognosis. Method: We studied 113 major Beta thalassemic patients who have been transplanted from 1990- 2000 in bone marrow transplantation center of Shariati Hospital. 62 were male and 51 were female. 27 pa¬tients were class one, 56 were class two and 30 were class three. The median age of each class were 6.5, 6.3 and 8.7. Conditioning regimen consisted of busulfan (3.5-4mg/Kg) and cyclophophamide (40-50mg/Kg).For GVHD prophylaxis we gave cyclosporine ± metothoroxate. Grade of liver fibrosis de¬fined by biopsy in all patients before BMT. All patients and their donors tested for HBSAg, HBSAb, HCVAb, CMVAb with RIA method. We assessed causes of liver dysfunction before and after trans¬plantation and effect of high ferritin level on liver function."nResults: Hepatic dysfunction in first year after transplantation was seen in 86 (76%) patients. Causes of liver dysfunction were consisted of 53.1% GVHD, 15.93% cyclosporine hepatotoxicity, 7.07% condi¬tioning regimen hepatotoxicity and VOD. In all three classes hepatic GVHD, cyclosporine toxicity, death and normal liver function post BMT had significant relation with hepatic dysfunction before BMT (P=0.001). In patients with ferritin level more than 1000, there were significant hepatotoxicity with conditioning regimen (P=0.001). 17 (15.04%) of patients have been died. Discussion: According to our study hepatic GVHD (%53.1) is the most common cause of hepatic dys¬function in all three classes.

Published
2005

27. Assessment of in vitro aging of mesenchymal stem cell

Author

Mohyeddin Bonab M, Alimoghaddam K, Talebian F, Ghaffari S.H, Ghavamzadeh A, and Nikbin B

Subjects
Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Mesenchyml stem cell (MSC) are receiving much attention in treatment of various diseases. The low frequency of MSCs in bone marrow (BM) necessitates their in vitro expansion prior to clinical use. We evaluated the effect of long term culture on the senescence of these cells."nBM cells were taken from 11 transplant donors with mean age of 25 years. In different passages, MSC were examined for different aging indicators including: telomere length assay, differentiation ability, immunophenotyping of CD13, CD44 and CD34 antigens, determination of cumulative population dou¬blings (CPDs), and study of morphological characteristics of MSC cultures."nThe mean long term culture was 118 day and the mean passage number was 9. The average number of PD decreased from 7.7 to 1.2 in the 10th passage. The mean telomere length decreased from 9.19 Kbp to 8.7 kbp in the 9th passage. Differentiation potential dropped from the 6th passage on. The culture's morphological abnormalities were typical of the Hayflick model of cellular aging. We believe that MSC enter senescence almost undetectably from the moment of in vitro culturing. Si¬multaneously these cells are losing their stem cell characteristics. Therefore, it is much better to con¬sider them for cell and gene therapy early on.

Published
2005

28. Phase II study of Gemcitabine and Cisplatin Regimen in Advanced Non-Small Cell Lung Cancer (NSCLC).

Author

Hoseinzadeh Mollayosefy M, Iravani M, Ghavamzadeh A, Toogheh Gh, and Alimoghaddam K

Subjects
Non small cell lung cancer (NSCLC), .Gemcitabine, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Background: Cisplatin-based chemotherapy is the standard treatment for advanced non-small cell lung cancer (NSCLC). Many novel drugs, including gemcitabine, vinorelbine, paclitaxel and docetaxel have been used in combination with cisplatin in this setting. Of these drugs, gemcitabine is reported to have a high response rate and acceptable toxicity. The aim of this study was to evaluate the efficacy and safety of gemcitabine & cisplatin combination."nMethods: Twenty-three patients with NSCLC, who met inclusion criteria, were enrolled from January 2001 till September 2003. All of them were confirmed by histology and were in advanced stages, i.e. stage IIIB or stage IV. Cisplatin with a dose of 70mg/m2 was given every 21 days, in combination with gemcitabine at a dose of 1250mg/m2 administered on days 1and 8 of a 21-day cycle. Results: of the 23 patients, 1 showed complete remission, 5 achieved partial remission, 7 had stable disease and 2 patients showed progressive disease, while 8 patients were not evaluable for response. The overall response in 15 evaluable patients was 40% (95% CI), median survival was 13.5 months (95% CI, 3.5-27.4 months), and median progression free survival (PFS)was 11 months (95% CI, 1.04-20.9 months)."nHematological toxicities included WHO grade 3, 4 anemia, neutropenia and thrombocytopenia 10%, 7% and 2% respectively. Non-Hematological toxicities included nausea/vomiting WHO grade 1,2 & peripheral neuropathy WHO grade 1,2. Skin rashes were mild.Six patients developed grade 2 toxicity. Renal impairment was mild. One case developed Acute Respiratory distress syndrome (ARDS) after first dose of chemotherapy, another case developed transient acute psychosis under therapy. Conclusions: The regimen of combined gemcitabine with cisplatin is safe and effective and well toler¬ated in patients. Some rare but important toxicity such as ARDS may occur occasionally. In this com¬bination, a lower dose of cisplatin seems to have an efficacy similar to that of in previous reports.

Published
2005

29. Imipenem/Cilastatin versus Cefepime as Empiric Monotherapy for Fever in Neutropenic Patients after ematopoietic Stem Cell Transplan¬tation

Author

Kani C., Mousavi A., Iravani M, Alimoghaddam K, Bahar B, Jahani M, and Ghavamzadeh A

Subjects
neutropenia Monotherapy Hematopoietic stem cell transplantation Randomized trial, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Objective: To evaluate the potential advantages of imipenem/cilastatin in control of fever in neutro-penic HSCT recipients.Patients and Method: In this single-center study, 111 consecutive febrile episodes in 104 neutropenic HSCT recipients with a mean age of 26 years were randomized to treatment either with Imipenem/cilastatin 1 g, IV, q8h or cefepime (our standard regimen) 2 g, IV, q8h. If fever persisted, se¬quential antibiotics were added in 72-hour intervals: vancomycin, amikacin and amphotericin-B. The study population was at serious risk of a poor outcome, since 73.5% of febrile episodes occurred after allogeneic and 26.5% of febrile events occurred after autologous hematopoietic stem cell transplanta¬tion."nResults: The median total duration of neutropenia was 10 days, and the median leukocyte count at study inclusion was 0.16 × 109/l. The two patient groups were comparable in terms of Age, gender, un¬derlying disease, conditioning regimen, clinical and bacterial documentation, severity and duration of neutropenia and mucositis, GI decontamination and G-CSF administration. Bacteremia was found in 20.6%, other microscopically documented infections in 9.8%, clinically documented infections in 20.6% and fever of unknown origin in 49% of the febrile episodes. Most (102) febrile episodes were evaluable for response. No significant difference was found between imipenem/cilastatin and cefepime in terms of success rate (73.1% versus 62%), empirical addition of vancomycin (38% versus 26.2%) or median duration of antibiotic therapy (7 days in both).The difference between imipenem/cilastatin and cefepime was statistically significant for median duration of fever (1.5 versus 2 days) and median time of resolution of neutropenia (12 versus 14 days). The overall response rates to initial monotherapy was significantly higher for HSCT recipients with thalassemia, MM, lymphoma, AA, than recipients with ALL, AML, CML, CLL (P

Published
2005

30. Arsenic trioxide induce apoptosis independent of TNFR-I and CD30 pathways in Acute promyelocytic leukemia patient with t(15;17) translocation.

Author

Ardjmand.AR, Alimoghadam. K, Kaviani S, Ghavamzadeh A, Djahani.M, and Moezzi L

Subjects
CD30, TNFR-I, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Arsenic trioxide (ATO) has been reported to induce apoptosis in Leukemic cells of Acute Promyelo-cytic Leukemia (APL) patients through different pathways. However, the exact mechanism of ATO-induced apoptosis is not yet clear. Co stimulation of death receptors CD30 and tumor necrosis factor receptor type one (TNFR-I) is one of the postulated mechanisms.In the present study we aimed to evaluate their involvement in fresh Promyelocytic cells separated from bone marrow of APL patients. Immunomagnetic separated cells were treated up to 48 hr at clinically tolerable concentrations of ATO (0.5-2.0 µmol/l) and expression of TNFR-I and CD30 were evaluated within the apoptotic and live populations using a sensitive triple color flow cytometric method for measuring apoptosis in combina¬tion with dual color immunofluorescence."nOur results suggest that the expression of TNFR-I and CD30 might not be related to ATO-induced apoptotic cell death.

Published
2005

31. In Search of Mesenchymal Stem Cells: Bone Marrow, Cord Blood, or Peripheral Blood

Author

Mohyeddin Bonab M, Alimoghaddam K, Talebian F, Ghaffari SH, Ghavamzadeh A, and Nikbin B.

Subjects
Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Objective: Mesenchymal stem cells (MSC) are capable of self-renewal and differentiation into various connective tissue lineages. Therefore, they have attracted a lot of attention from investigators in the context of stem cell therapies. In our study, we have evaluated the frequency, phenotype and differen¬tiation potential of MSC in bone marrow (BM) cord blood (CB) and mobilized peripheral blood (mPB). methods: Sixteen CB, 11 BM and 19 mPB were obtained from normal donors. Mononuclear cells sus¬pended in culture medium and seeded in culture flasks. Flasks were incubated in a CO2 incubator with a change of culture medium every 4 days and passaged when fibroblast like cells reached confluence. For every other passage, MSC were examined for CD13, CD44,CD34 by flow cytometry and induced to differentiate into adipocytes and osteocytes."nResults: All BM samples produced MSC that survived multiple passages in mesenchymal culture me¬dium over 4 months. CB and mPB samples produced a non-confluent adherent layer of heterogeneous cells, and did not proliferate beyond the first passage. Immunophenotype of BM-derived MSC in every other passage were CD34-, CD13+ and CD44+, the adipogenic and osteogenic differentiation were con¬firmed by Oil-red O and Von Kossas staining, respectively."nThe mentioned evaluation for mPB and CB were not attempted because these were not confluent even in the first passage."nConclusion: In our study, only human BM cells produced MSC. These cells are positive for MSC sur¬face proteins and differentiate into MSC lineages.

Published
2005

32. Evaluation Of Angiogenesis In The Bone Marrow Of Patients With Acute Myloid Leukemia

Author

Sanaat Z, Tavangar M, Shriftabrizi A, Alimoghadam K, Ghavamzadeh A, and Jahani M

Subjects
AML, angiogenesis, microvascular density, Medicine (General), R5-920
Abstract

Background: The important of angiogenesis for the progressive growth and viability of solid tumors is well established. Only few data are available for hematologic neoplasms. Materials and Methods: To investigate the role of angiogenesis in the acute myloid leukemia (AML) bone marrow biopsies from 30 adults with newly diagnosed, untreated AML(day 0) were evaluated. Further studies were done after completion on remission induction of treatment (day 35 of 7×3 regimen n=13, complete remission in AML (m3) treat with arsenic trioxide n=17). Micro-vessels were scored in at least 3 areas of highest micro-vessel density in representative section of each bone marrow specimen using immunohistochemistry for Von Willbrand factor. Results: Median micro-vascular density (MVD) were in AMLM3 patients before treatment, %6.81±3.58 and after treatetment %3.48±3.06 (p

Published
2004

33. Robust Reinforcement Learning using Offline Data

Author

Panaganti, Kishan, Xu, Zaiyan, Kalathil, Dileep, and Ghavamzadeh, Mohammad

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Statistics - Machine Learning
Abstract

The goal of robust reinforcement learning (RL) is to learn a policy that is robust against the uncertainty in model parameters. Parameter uncertainty commonly occurs in many real-world RL applications due to simulator modeling errors, changes in the real-world system dynamics over time, and adversarial disturbances. Robust RL is typically formulated as a max-min problem, where the objective is to learn the policy that maximizes the value against the worst possible models that lie in an uncertainty set. In this work, we propose a robust RL algorithm called Robust Fitted Q-Iteration (RFQI), which uses only an offline dataset to learn the optimal robust policy. Robust RL with offline data is significantly more challenging than its non-robust counterpart because of the minimization over all models present in the robust Bellman operator. This poses challenges in offline data collection, optimization over the models, and unbiased estimation. In this work, we propose a systematic approach to overcome these challenges, resulting in our RFQI algorithm. We prove that RFQI learns a near-optimal robust policy under standard assumptions and demonstrate its superior performance on standard benchmark problems., Comment: Appeared in Neural Information Processing Systems (NeurIPS) 2022

Published
2022

34. Non-Myeloablative Stem Cell Transplantation in Hematologic Malig¬nancies: An Experience from the Hematology-Oncology and BMT Re¬search Center

Author

Keyhanian S, Ghavamzadeh A, Bahar B, Alimoghaddam K, Shamshiri AR, and Gholibeikian.S

Subjects
Hematopoietic stem cell transplantation, Allogeneic, Non-myeloablative, Graft vs Host Dis¬ease, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Myeloablative-allogeneic stem cell transplantation is a common way of treating various malignant and nonma-lignant diseases; but, it is associated with hazardous immediate and late complications. The majority of patients are not good candidates for high dose therapy because of old age, medical co-morbidities or previous heavy treatments. The donor stem cells can engraft in the recipient and induce mixed chimerism when we use a less intensive, but sufficiently immunosup-pressive, conditioning regimen, known as mini-transplantation or non-Myeloablative allogeneic Stem Cell Transplantation (NM-allo-SCT)."nMethods: The conditioning regimens were the combination of Fludarabine and Cyclophosphamide or Busulfan and ATG. Prophylaxis against graft versus host disease (GVHD) included Cyclosporine A (CSA) +/- Methotrexate. A multiplex-PCR using short tandem repeats (VNTR) was used for chimerism analysis."nResults: We report the results of NM-allo-SCT from the HLA-identical siblings in 20 patients with AML (N=7), CML (N=6), NHL (N=2), MDS (N=2), ALL (N=1) and Fanconi anemia (N=2). Fourteen males and 6 females with median age of 43 years (range 8-55) underwent NM-allo-SCT and were followed up 4-870 days (median 420 days). Typical side effect of conventional HSCT, such as severe mucositis, vomiting and VOD were absent. Most of the patients did not become se¬verely pancytopenic and had relatively short hospitalization. Hematological recovery was rapid, a median of 8.5 days. Acute GVHD (grade ≥II) and extensive chronic GVHD was observed in three patients. Most of the patients initially had mixed-chimerism, progressing to full-donor-chimerism in 11 patients, after the interruption of the CSA therapy, and, in one patient, after DLI. Nine patients died, six from relapse or disease progression and three from transplantation-related complications (GVHD, infection or secondary malignancy). 14 month overall survival and disease free survival of 55% and 50%, respec¬tively, was observed."nConclusion: Our results confirm that NM-allo-SCT is safe and minimally toxic and is a potential new approach for a safer treatment of a large variety of hematologic diseases, especially in patients with AML and CML in remission.

Published
2004

35. The Role Of Interleukin - 18 And Interleukin – 2 Receptors In Acute Graft-Versus-Host Disease After Bone Marrow Transplantation

Author

Iravani M, Shayegan M, Babaei G, Talebian A, Ghavamzadeh A, Babak Bahar, and Aghaeipoor M

Subjects
IL-18, sIL–2R, aGVHD, BMT, Medicine (General), R5-920
Abstract

Background: Graft-versus-host disease is one of the major complications after allogenic bone marrow transplantation, but it is not easy to anticipate the onset. Cytokines released by type 1 T-helper cells are thought to play a pivotal role in acute graft-versus-host disease (aGVHD). The ability to predict the likely occurrence of graft-versus-host-disease (GVHD) after BMT would be extremely valuable. By serially measuring serum levels of soluble IL-2 receptor (sIL-2R), IL-18 and following allogeneic bone marrow transplantation (BMT), we tried to define their relationship to aGVHD as complication of the transplantation and determine useful markers for aGVHD predictors. Materials and Methods: Serum sIL-2R, IL-18, and levels were measured by sandwich ELISA in 219 sera samples from 39 patients (with hematological disorders before and after allogeneic BMT) and 28 controls. All patients received BMT from HLA-identical siblings. Results: 25 patients developed aGVHD and serum levels of sIL-2 R and IL-18 , in sera drawn before transplantation , in patients with acute graft-versus-host disease (aGVHD +) , were increased in comparison of patients without acute graft-versus-host disease (aGVHD ¯) and control group and there wasn’t any significant differences in serum levels of sIL-2 R and IL-18 in aGVHD ¯ patients and controls. Serum level of IL-18, in aGVHD+ patients, was increased during day 3 - 24 after BMT, and there was a significant difference in patients with GVHD 0 – GVHD III. In majority of patients with acute GVHD (60 %) , the peak levels of IL-18 and IL-2R was achieved on day 10 after BMT and the rise in sIL-2R and IL-18 preceded of clinical signs of GVHD (mean day 15 after BMT). Level of IL-18 in patients with aGVHD had strongly correlated with the severity of aGVHD on Day 10 after BMT. IL-18 level mean (before BMT), in patients who received Busulfan and Fludarabin to treat aGVHD, was lower than in patients who received Busulfan - Endoxan, or Cyclophosphamide. Conclusion: Our data concluded that IL-18 plays an important role in the development of aGVHD and IL-18 level might be an indicator for aGVHD, reflecting the severity of the disease. These findings suggest that IL-18 may play important roles in the pathogenesis of aGVHD and that measurement of serum IL-18 levels can be useful predictor of aGVHD.

Published
2004

36. Arsenic Trioxide Selectively Induces Apoptosis within the Leukemic Cells of APL Patients with t(15;17) Translocation Possibly through the Fas Pathway

Author

Ardjmand. A.R, Alimoghaddam. K, Zaker F, Ghavamzadeh A, and Jahani.M

Subjects
Acute Promyelocytic Leukemia, Fas/Apo1, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Acute Promyelocytic Leukemia is a sub-type of acute myelogenous leukemia that occurs in about 10-15% of patients with AML. Approximately 20%-30% of these patients, who are treated with the current standard All Trans Retinoic Acid (ATRA) and Anthracyclin-based chemotherapy regimen, suffer relapse in less than a year. Arsenic trioxide (ATO) as a single agent can induce complete remission even in refractory and relapsed patients with few adverse effects. The investiga¬tors efforts regarding elucidation of the mechanisms of action underlying these clinical responses has shown that Arsenic apparently affects numerous intracellular signal transduction pathways and causes many alterations in cellular function, among which the most prominent ones are the induction of differentiation & apoptosis with low & high doses of arsenic, respectively."nPurposes: In vivo apoptosis on these patients has not been evaluated yet and despite previous In vitro studies, which mostly reveal Fas/Apo1 is not expressed during ATO treatment, its in vivo expression has not been evaluated yet. Materials & methodes: In order to study the apoptotic pattern in leukemic cells of APL patients, we conducted a single-laser, triple-color flowcytometric experiment, to detect leukemic apoptotic cells in a heterogeneous population of bone mar¬row samples with the Annexin V & 7AAD technique. The Fas expression was also evaluated in promyelocyte population cells in a dual color panel."nResults & Conclusion: A substantial Apoptosis was selectively detected in Promyelocytic cells during the early and middle stages of treatment and the concurrent Fas expression indicates its involvent in Apoptosis induced by Arsenic Trioxide.

Published
2004

37. Reinforcement Learning of Multi-Domain Dialog Policies Via Action Embeddings

Author

Mendez, Jorge A., Geramifard, Alborz, Ghavamzadeh, Mohammad, and Liu, Bing

Subjects
Computer Science - Computation and Language, Computer Science - Machine Learning
Abstract

Learning task-oriented dialog policies via reinforcement learning typically requires large amounts of interaction with users, which in practice renders such methods unusable for real-world applications. In order to reduce the data requirements, we propose to leverage data from across different dialog domains, thereby reducing the amount of data required from each given domain. In particular, we propose to learn domain-agnostic action embeddings, which capture general-purpose structure that informs the system how to act given the current dialog context, and are then specialized to a specific domain. We show how this approach is capable of learning with significantly less interaction with users, with a reduction of 35% in the number of dialogs required to learn, and to a higher level of proficiency than training separate policies for each domain on a set of simulated domains., Comment: Presented in the Conversational AI Workshop, NeurIPS 2019

Published
2022

38. Neoadjuvant and Adjuvant Chemotherapy in Osteosarcoma (The Experience of HORC in the Shariati Hospital)

Author

Alimoghadam K, Ghavamzadeh A, Jafari M., Sami Hagialilo S, Jahani M, khodabandeh A, Eghbal L, and Gholibekian S

Subjects
Neoadjuvant chemotherapy, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Osteosarcoma is the most common bone sarcoma, and the third most common malignancy in children and adolescents. Before 1970, amputation was the sole treatment. Eighty percent of patients died from metastatic diseases, most commonly in the lungs. Over the past three decades, effective neoadjuvant (preoperative) and adjuvant (postoperative) chemotherapy protocols have improved the ability to perform limb salvage resections, disease free survival and overall survival rates."nPatients and Methods: The study was conducted on 28 patients (15 male and 13 female) whose diagnoses were confirmed by excisional biopsy without any proof of metastasis in clinical and radiological assessments from September 2001 to November 2002. All patients were treated with three-drug regimen consisting of Adriamycin, Ifosfamide and Cisplatin. The neoadjuvant chemotherapy was administered in three courses. The first course, Ifosfamide (2gr/IV) and Adriamycin (75mg/m², IV infusion) were given on the first day and Ifosfamide (1.5 gr/m² by continuous infusion) alone for 6 days. The second course consisted of Adriamycin (75mg/m², IV infusion) and Cisplatin (100mg/m², IV infusion) for one day. The third course was the same as the first. After surgery, all patients received adjuvant therapy similar to the neoadjuvant protocol mentioned above. Limb salvage was the most common surgical method. The treatment outcome particularly depended on the percentage of tumor necrosis. Overall and disease-free survival were also measured."nResults: According to the tumor necrosis percentage, the tumor response to chemotherapy was classified from good to poor response. In this study, 63.6% of patients showed good response and 36.4% indicated poor or no response to chemotherapy. The tumor necrosis percentage was significantly correlated with age≤ 20 years (P= 0.01), tumor size ≤84 cm³ (P= 0.03) and the site of tumors in femurs (P= 0.03). The average follow-up time was 132 days, ranging from 15 to 618 days. The first year survival rate was 100%, and the disease-free survival (DFS) was 70.8% for the same time period. Disease-free survival was significantly correlated with the chemotherapy response (P= 0.03), which was 100% in the good response group in the first year."nConclusion: Although we had utilized bone grafts for substantially resected bones, local relapses were remarkably low (2 cases), so we suggest that this surgical method can be a proper alternative treatment for different types of expensive prosthesis in countries with low socioeconomic status.

Published
2004

39. Chronic Graft versus Host Disease after Allogeneic Bone Marrow Transplantation; An Analysis of Incidence and Risk Factors.

Author

Ghavamzadeh A, Alimoghaddam K, Bahar B, and Foroughi F

Subjects
and Complications, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Chronic graft versus host disease (cGVHD) is one of the most serious potential complications of"nallogeneic bone marrow transplantation."nStudy design and method: We analyzed the incidence of cGVHD and its associated risk factors in a group of 161"nIranian recipients of HLA-identical sibling transplants, with at least 90 days post-transplantation survival. In the"nmajority of cases (n=73), cGVHD occurred in the first year after the transplant (median 273 days). The actual"nprobability of cGVHD within 1 year was 45.3±7% (CI 95%)."nResults: In a univariate analysis, the most important risk factor was the type of transplant. Peripheral blood stem cell"ntransplants (PBSCT) showed a significant increase in cGVHD compared with bone marrow transplants (BMT)"n(RR=2.34, pBM, p

Published
2004

40. Is There any Greater Possibility in Finding HLA-identical Unrelated Hematopoietic Stem Cell Donors among Thalassemia Families for Transplantation of Thalassemia Patients?

Author

Mohyeddin M, Alijanipour P, Alimoghaddam K, Ghavamzadeh A, Khosravi F, and Nikbin B.

Subjects
HLA, Stem cell, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Thalassemia is probably the most common single gene disorder causing a major public health problem in the world. Currently, allogenic hematopoietic stem cell transplantation (HSCT) is the only curative therapy for thalassemia. One major limitation of HSCT is the lack of HLA-identical sibling donors, so attention has turned to finding phenotypically matched unrelated donors."nPatients and methods: From 1991 to 2002, 182 thalassemia patients referred to our center for HSCT. Donor selection was based on HLA class I and class II histocompatibility matching. The results of the serologically HLA class I typing of 549 subjects (patients and their families) and HLA class II typing of 182 patients were compared with HLA class I and II antigens of 100 healthy Iranians normal people. The comparisons between these two groups were tested in univariate analysis, using the Pearson chi-squared statistics."nResults: In comparing, thalassemic families (549 subjects) and healthy Iranians (100 subjects) for HLA class I antigens, significant differences for 11 antigens, including A9 (p= 0.029), A11 (p= 0.01), A19 (p= 0.000), B16 (p= 0.000), B17 (p= 0.029), B27 (p= 0.003), B41(p= 0.000), C2 (p= 0.015), C3(p= 0.012), C4 (p= 0.004), C7 (p= 0.000) were found. For HLA class II antigens, we found that only HLA-DR7 was significantly different (p= 0.002) between 182 thalas-semia patients and the healthy Iranian normal group."nConclusion: In this study, we found that thalassemia families showed significant differences, compared to the healthy Iranian group in several HLA antigens. Comparing HLA polymorphism and finding enough similarity in thalassemia families in the countries, located in the thalassemia belt, may provide benefits for establishing a common HLA bank of thalassemia families.

Published
2004

41. Gastrointestinal Bleeding, the First Presentation of Acute GVHD in Two Patients with Thalassemia after BMT

Author

Ghavamzadeh A, Moosavi A, Hedayatiasl A, and Taghipour R

Subjects
Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Engraftment by donor lymphocytes in an immunologically compromised host can result in donor T-cell activation against host major histocompatibility complex antigens, with resultant GVHD."nThe acute form of GVHD (aGVHD) is characterized by erythroderma, cholestatic hepatitis, and enteritis. The intestinal symptoms of aGVHD include crampy abdominal pain and watery diarrhea, often with blood. The conditioning regimen and infectious agents may produce similar symptoms. Severe intestinal aGVHD is a life-threatening event and associated with high mortality."nIn this case report, we describe two patients with major thalassemia who experienced acute gastrointestinal GVHD. One of them experienced it after peripheral blood transplantation at day +13, and the other after bone marrow transplantation at day +14. The first presentation was severe GI bleeding, and then 2 litters per day diarrhea. Besides standard prophylaxis with Cyclosporine and Methotrexate, Methylprednisolone 2mg/kg per day commenced because GI bleeding started and afterward supportive treatment means were continued. Following administration of Methylprednisolone, the amount of GI bleeding and diarrhea declined; in addition, the need for whole blood transfusion and blood products decreased. Both children had no problem in follow-up."nEngraftment evaluation by the VNTR method showed 100 percent validity. GI bleeding after transplantation can be a major presentation of aGVHD, which requires precise attention, and on-time treatment. The elimination of other causes of GI bleeding and diarrhea, in addition to the two other factors mentioned above, would increase the survival rate of patients greatly.

Published
2004

42. Evaluation of the Affect of Maternal and Neonatal Factors on Cord Blood Parameters?

Author

Mohyeddin Bonab M, Goliaei Z, Alimoghaddam K, and Ghavamzadeh A

Subjects
nucleated cell, neonatal factors, maternal factors, Neoplasms. Tumors. Oncology. Including cancer and carcinogens, RC254-282
Abstract

Chronic graft versus host disease (cGVHD) is one of the most serious potential complications of"nallogeneic bone marrow transplantation."nStudy design and method: We analyzed the incidence of cGVHD and its associated risk factors in a group of 161"nIranian recipients of HLA-identical sibling transplants, with at least 90 days post-transplantation survival. In the"nmajority of cases (n=73), cGVHD occurred in the first year after the transplant (median 273 days). The actual"nprobability of cGVHD within 1 year was 45.3±7% (CI 95%)."nResults: In a univariate analysis, the most important risk factor was the type of transplant. Peripheral blood stem cell"ntransplants (PBSCT) showed a significant increase in cGVHD compared with bone marrow transplants (BMT)"n(RR=2.34, pBM, p

Published
2004

43. 'Time sequential high dose of Cytarabine in acute myelocytic leukemia '

Author

Ghavamzadeh A, Jahani M, Alimoghaddam K, Iravani M, Aghadmi N, and Tavassoli P

Subjects
Acute myelocytic leukemia, Antineoplastic protocols, High dose cytarabine, Medicine (General), R5-920
Abstract

Given preliminary evidence of timed, sequential chemotherapy of high dose cytosine arabinoside the current study was initiated to assess the side effects and efficacy of this regimen in patients with newly acute myelocytic leukemia (AML). Nineteen adults who referred to Hematology-Oncology and Bone Marrow Transplantation (BMT) research center of Tehran University of Medical Sciences were enrolled in a trial from Aug 1999 to Nov 2000. All patients had a Karnofski classification above 60%. At this time induction therapy consisted of daunorubicin or idarubicin given at a dose of 60 mg/m² and 12 mg/m² IV respectively on days 1-3, and cytarabine (Ara-C) 100 mg/m² intravenously by continuous infusion on days 1-7, followed by Ara-C 1000 mg/m² given on day 8-10 every 12 hours by IV infusion. Consolidation therapy started after 35th day. Of 19 fully evaluable patients, 10 patients achieved a complete remission, whereas 36.6% patients succumbed to death due to regeneration failure. The clinical data show that the overall survival rate from diagnosis 55.5% (95% CI, 30.8-78.5) at 6 months for the entire cohort of the patients. Disease free survival is also 50% (95% CI, 26-74). Mean duration of death due to treatment was 20 days (range 17-29) after beginning the regimen. Presenting WBC counts, French-American-British (FAB) classification, sex and age were not useful prognostic variables. Fever, diarrhea, nausea and vomiting and GI hemorrhage were seen in 19, 6, 4, 7 patients respectively. It seems the 3+7+3 regimen is a promising approach for the AML patients regarding to high complete remission rate, but more supportive care should be considered. Furthermore any, benefit in long-term outcome can’t be determined regardless to the choice of post remission therapy (e.g., GCSF, appropriate antibiotics and etc).

Published
2003

44. 'Transformation of chronic myelogenous leukemia to Multiple Myeloma: A case report '

Author

Ghavamzadeh A, Alimoghaddam K, Mehdipour P, Sharifian R, Schwanitz G, and Shamshiri AR "

Subjects
Myeloid, Chronic, Cell transformation, Neoplastic, Philadelphia chromosome, Medicine (General), R5-920
Abstract

Chronic myelogenous leukemia (CML) is a stem cell disorder sometimes associated with lymphoproliferative disorders. CML may precede a lymphoproliferative disorder. There are a few reports showing associating of CML with multiple myeloma and we report a known CML case that transformed into a full-blown multiple myeloma. This patient had more than 69% of infiltrating myeloma cells in her bone marrow and Philadelphia chromosome was detected in 18 out of 42. However, the probable presence of some myeloma cells with classic Philadelphia-positive chromosome could be proposed.

Published
2003

45. 'Thalassemia: Incidence and predictive factors for chronic GVHD after HLA-identical sibling marrow transplantation '

Author

Ghavamzadeh A, Alimoghadam K, Bahar B, Foroughi F, and Jahani M "

Subjects
Chronic GVHD, Medicine (General), R5-920
Abstract

Allogeneic bone marrow transplantation is the only definie cure in thalassemia and its most important complication is chronic graft-versus-host disease (cGVHD). We analysed the incidence of cGVHD and its associated risk factors in a group of 89 Iranian thalassemic patients of HLA-identical sibling transplants surviving at least 90 days after transplantation.In the majority of cases (39) cGVHD occurred in the first year following transplant (median 271 days). Actuarial probability of cGVHD in 1 year was 43.8±10% (95% CI). In univariate analysis, the most important risk factor was the type of transplant: 78.9% (15.19) of patients who underwent peripheral blood stem cell transplant developed cGVHD compared with only 34.3% (24/70) of those who underwent bone marrow transplant (RR=3.65 p BM p

Published
2002

46. A Mixture-of-Expert Approach to RL-based Dialogue Management

Author

Chow, Yinlam, Tulepbergenov, Aza, Nachum, Ofir, Ryu, MoonKyung, Ghavamzadeh, Mohammad, and Boutilier, Craig

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence
Abstract

Despite recent advancements in language models (LMs), their application to dialogue management (DM) problems and ability to carry on rich conversations remain a challenge. We use reinforcement learning (RL) to develop a dialogue agent that avoids being short-sighted (outputting generic utterances) and maximizes overall user satisfaction. Most existing RL approaches to DM train the agent at the word-level, and thus, have to deal with a combinatorially complex action space even for a medium-size vocabulary. As a result, they struggle to produce a successful and engaging dialogue even if they are warm-started with a pre-trained LM. To address this issue, we develop a RL-based DM using a novel mixture of expert language model (MoE-LM) that consists of (i) a LM capable of learning diverse semantics for conversation histories, (ii) a number of {\em specialized} LMs (or experts) capable of generating utterances corresponding to a particular attribute or personality, and (iii) a RL-based DM that performs dialogue planning with the utterances generated by the experts. Our MoE approach provides greater flexibility to generate sensible utterances with different intents and allows RL to focus on conversational-level DM. We compare it with SOTA baselines on open-domain dialogues and demonstrate its effectiveness both in terms of the diversity and sensibility of the generated utterances and the overall DM performance.

Published
2022

47. Collaborative Multi-agent Stochastic Linear Bandits

Author

Moradipari, Ahmadreza, Ghavamzadeh, Mohammad, and Alizadeh, Mahnoosh

Subjects
Computer Science - Machine Learning, Computer Science - Multiagent Systems
Abstract

We study a collaborative multi-agent stochastic linear bandit setting, where $N$ agents that form a network communicate locally to minimize their overall regret. In this setting, each agent has its own linear bandit problem (its own reward parameter) and the goal is to select the best global action w.r.t. the average of their reward parameters. At each round, each agent proposes an action, and one action is randomly selected and played as the network action. All the agents observe the corresponding rewards of the played actions and use an accelerated consensus procedure to compute an estimate of the average of the rewards obtained by all the agents. We propose a distributed upper confidence bound (UCB) algorithm and prove a high probability bound on its $T$-round regret in which we include a linear growth of regret associated with each communication round. Our regret bound is of order $\mathcal{O}\Big(\sqrt{\frac{T}{N \log(1/|\lambda_2|)}}\cdot (\log T)^2\Big)$, where $\lambda_2$ is the second largest (in absolute value) eigenvalue of the communication matrix.

Published
2022

48. Multi-Environment Meta-Learning in Stochastic Linear Bandits

Author

Moradipari, Ahmadreza, Ghavamzadeh, Mohammad, Rajabzadeh, Taha, Thrampoulidis, Christos, and Alizadeh, Mahnoosh

Subjects
Computer Science - Machine Learning
Abstract

In this work we investigate meta-learning (or learning-to-learn) approaches in multi-task linear stochastic bandit problems that can originate from multiple environments. Inspired by the work of [1] on meta-learning in a sequence of linear bandit problems whose parameters are sampled from a single distribution (i.e., a single environment), here we consider the feasibility of meta-learning when task parameters are drawn from a mixture distribution instead. For this problem, we propose a regularized version of the OFUL algorithm that, when trained on tasks with labeled environments, achieves low regret on a new task without requiring knowledge of the environment from which the new task originates. Specifically, our regret bound for the new algorithm captures the effect of environment misclassification and highlights the benefits over learning each task separately or meta-learning without recognition of the distinct mixture components.

Published
2022

49. Efficient Risk-Averse Reinforcement Learning

Author

Greenberg, Ido, Chow, Yinlam, Ghavamzadeh, Mohammad, and Mannor, Shie

Subjects
Computer Science - Machine Learning
Abstract

In risk-averse reinforcement learning (RL), the goal is to optimize some risk measure of the returns. A risk measure often focuses on the worst returns out of the agent's experience. As a result, standard methods for risk-averse RL often ignore high-return strategies. We prove that under certain conditions this inevitably leads to a local-optimum barrier, and propose a soft risk mechanism to bypass it. We also devise a novel Cross Entropy module for risk sampling, which (1) preserves risk aversion despite the soft risk; (2) independently improves sample efficiency. By separating the risk aversion of the sampler and the optimizer, we can sample episodes with poor conditions, yet optimize with respect to successful strategies. We combine these two concepts in CeSoR - Cross-entropy Soft-Risk optimization algorithm - which can be applied on top of any risk-averse policy gradient (PG) method. We demonstrate improved risk aversion in maze navigation, autonomous driving, and resource allocation benchmarks, including in scenarios where standard risk-averse PG completely fails., Comment: Accepted to NeurIPS 2022

Published
2022

50. Non-stationary Bandits and Meta-Learning with a Small Set of Optimal Arms

Author

Azizi, MohammadJavad, Duong, Thang, Abbasi-Yadkori, Yasin, György, András, Vernade, Claire, and Ghavamzadeh, Mohammad

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

We study a sequential decision problem where the learner faces a sequence of $K$-armed bandit tasks. The task boundaries might be known (the bandit meta-learning setting), or unknown (the non-stationary bandit setting). For a given integer $M\le K$, the learner aims to compete with the best subset of arms of size $M$. We design an algorithm based on a reduction to bandit submodular maximization, and show that, for $T$ rounds comprised of $N$ tasks, in the regime of large number of tasks and small number of optimal arms $M$, its regret in both settings is smaller than the simple baseline of $\tilde{O}(\sqrt{KNT})$ that can be obtained by using standard algorithms designed for non-stationary bandit problems. For the bandit meta-learning problem with fixed task length $\tau$, we show that the regret of the algorithm is bounded as $\tilde{O}(NM\sqrt{M \tau}+N^{2/3}M\tau)$. Under additional assumptions on the identifiability of the optimal arms in each task, we show a bandit meta-learning algorithm with an improved $\tilde{O}(N\sqrt{M \tau}+N^{1/2}\sqrt{M K \tau})$ regret.

Published
2022

Catalog

Books, media, physical & digital resources
See catalog results