Search

The two most commonly used methods to model the behaviour of granular flows are discrete element and continuum mechanics simulations. These approaches concentrate on the deterministic description of particle or bulk material motion. Unlike these approaches, this paper introduces an alternative model that describes the stochastic dynamics of the void spaces under the action of gravity. The model includes several key phenomena which are observed in deforming granular media, such as segregation, mixing, and an angle of repose. These mechanisms are modelled heterarchically using both spatial and microstructural internal coordinates. Key aspects of the model include its ability to describe both stable and flowing states of granular media based on a solid fraction cut-off, and the influence of particle size on flow, segregation, and mixing. The model is validated with simulations of column collapse and silo discharge.

Data scarcity in medical imaging poses significant challenges due to privacy concerns. Diffusion models, a recent generative modeling technique, offer a potential solution by generating synthetic and realistic data. However, questions remain about the performance of convolutional neural network (CNN) models on original and synthetic datasets. If diffusion-generated samples can help CNN models perform comparably to those trained on original datasets, reliance on patient-specific data for training CNNs might be reduced. In this study, we investigated the effectiveness of diffusion models for generating synthetic medical images to train CNNs in three domains: Brain Tumor MRI, Acute Lymphoblastic Leukemia (ALL), and SARS-CoV-2 CT scans. A diffusion model was trained to generate synthetic datasets for each domain. Pre-trained CNN architectures were then trained on these synthetic datasets and evaluated on unseen real data. All three datasets achieved promising classification performance using CNNs trained on synthetic data. Local Interpretable Model-Agnostic Explanations (LIME) analysis revealed that the models focused on relevant image features for classification. This study demonstrates the potential of diffusion models to generate synthetic medical images for training CNNs in medical image analysis.

For a given ground cost, approximating the Monge optimal transport map that pushes forward a given probability measure onto another has become a staple in several modern machine learning algorithms. The fourth-order Ma-Trudinger-Wang (MTW) tensor associated with this ground cost function provides a notion of curvature in optimal transport. The non-negativity of this tensor plays a crucial role for establishing continuity for the Monge optimal transport map. It is, however, generally difficult to analytically verify this condition for any given ground cost. To expand the class of cost functions for which MTW non-negativity can be verified, we propose a provably correct computational approach which provides certificates of non-negativity for the MTW tensor using Sum-of-Squares (SOS) programming. We further show that our SOS technique can also be used to compute an inner approximation of the region where MTW non-negativity holds. We apply our proposed SOS programming method to several practical ground cost functions to approximate the regions of regularity of their corresponding optimal transport maps.

The rise of large language models (LLMs) has created a significant disparity: industrial research labs with their computational resources, expert teams, and advanced infrastructures, can effectively fine-tune LLMs, while individual developers and small organizations face barriers due to limited resources. In this paper, we aim to bridge this gap by presenting a comprehensive study on supervised fine-tuning of LLMs using instruction-tuning datasets spanning diverse knowledge domains and skills. We focus on small-sized LLMs (3B to 7B parameters) for their cost-efficiency and accessibility. We explore various training configurations and strategies across four open-source pre-trained models. We provide detailed documentation of these configurations, revealing findings that challenge several common training practices, including hyperparameter recommendations from TULU and phased training recommended by Orca. Key insights from our work include: (i) larger batch sizes paired with lower learning rates lead to improved model performance on benchmarks such as MMLU, MTBench, and Open LLM Leaderboard; (ii) early-stage training dynamics, such as lower gradient norms and higher loss values, are strong indicators of better final model performance, enabling early termination of sub-optimal runs and significant computational savings; (iii) through a thorough exploration of hyperparameters like warmup steps and learning rate schedules, we provide guidance for practitioners and find that certain simplifications do not compromise performance; and (iv) we observed no significant difference in performance between phased and stacked training strategies, but stacked training is simpler and more sample efficient. With these findings holding robustly across datasets and models, we hope this study serves as a guide for practitioners fine-tuning small LLMs and promotes a more inclusive environment for LLM research., Comment: 33 pages, 19 figures. Appendix included in submission. Submitted to ICLR 2025

Metrology, the science of measurement, plays a key role in Advanced Manufacturing (AM) to ensure quality control, process optimization, and predictive maintenance. However, it has often been overlooked in AM domains due to the current focus on automation and the complexity of integrated precise measurement systems. Over the years, Digital Twin (DT) technology in AM has gained much attention due to its potential to address these challenges through physical data integration and real-time monitoring, though its use in metrology remains limited. Taking this into account, this study proposes a novel framework, the Metrology and Manufacturing-Integrated Digital Twin (MM-DT), which focuses on data from two metrology tools, collected from Coordinate Measuring Machines (CMM) and FARO Arm devices. Throughout this process, we measured 20 manufacturing parts, with each part assessed twice under different temperature conditions. Using Ensemble Machine Learning methods, our proposed approach predicts measurement deviations accurately, achieving an R2 score of 0.91 and reducing the Root Mean Square Error (RMSE) to 1.59 micrometers. Our MM-DT framework demonstrates its efficiency by improving metrology processes and offers valuable insights for researchers and practitioners who aim to increase manufacturing precision and quality.

While textless Spoken Language Models (SLMs) have shown potential in end-to-end speech-to-speech modeling, they still lag behind text-based Large Language Models (LLMs) in terms of semantic coherence and relevance. This work introduces the Align-SLM framework, which leverages preference optimization inspired by Reinforcement Learning with AI Feedback (RLAIF) to enhance the semantic understanding of SLMs. Our approach generates multiple speech continuations from a given prompt and uses semantic metrics to create preference data for Direct Preference Optimization (DPO). We evaluate the framework using ZeroSpeech 2021 benchmarks for lexical and syntactic modeling, the spoken version of the StoryCloze dataset for semantic coherence, and other speech generation metrics, including the GPT4-o score and human evaluation. Experimental results show that our method achieves state-of-the-art performance for SLMs on most benchmarks, highlighting the importance of preference optimization to improve the semantics of SLMs.

The $H_2$ norm is a commonly used performance metric in the design of estimators. However, $H_2$-optimal estimation of most PDEs is complicated by the lack of state-space and transfer function representations. To address this problem, we re-characterize the $H_2$-norm in terms of a map from initial condition to output. We then leverage the Partial Integral Equation (PIE) state-space representation of systems of linear PDEs coupled with ODEs to recast this characterization of $H_2$ norm as a convex optimization problem defined in terms of Linear Partial Integral (LPI) inequalities. We then parameterize a class of PIE-based observers and formulate the associated $H_2$-optimal estimation problem. The optimal observer synthesis problem is then recast as an LPI, and the resulting observers are validated using numerical simulation.

Over the years, Digital Twin (DT) has become popular in Advanced Manufacturing (AM) due to its ability to improve production efficiency and quality. By creating virtual replicas of physical assets, DTs help in real-time monitoring, develop predictive models, and improve operational performance. However, integrating data from physical systems into reliable predictive models, particularly in precision measurement and failure prevention, is often challenging and less explored. This study introduces a Predictive Maintenance and Inspection Digital Twin (PMI-DT) framework with a focus on precision measurement and predictive quality assurance using 3D-printed 1''-4 ACME bolt, CyberGage 360 vision inspection system, SolidWorks, and Microsoft Azure. During this approach, dimensional inspection data is combined with fatigue test results to create a model for detecting failures. Using Machine Learning (ML) -- Random Forest and Decision Tree models -- the proposed approaches were able to predict bolt failure with real-time data 100% accurately. Our preliminary result shows Max Position (30%) and Max Load (24%) are the main factors that contribute to that failure. We expect the PMI-DT framework will reduce inspection time and improve predictive maintenance, ultimately giving manufacturers a practical way to boost product quality and reliability using DT in AM.

Solids subjected to repeated cycles of stress or deformation can fail after several cycles, a phenomenon termed fatigue failure. Although intensely investigated for a wide range of materials owing to its obvious practical importance, a microscopic understanding of the initiation of fatigue failure continues to be actively pursued, in particular for soft and amorphous materials. We investigate fatigue failure for glasses subjected to cyclic shear deformation through computer simulations. We show that, approaching the so-called fatigue limit, failure times display a power law divergence, at variance with commonly used functional forms, and exhibit strong dependence on the degree of annealing of the glasses. We explore several measures of damage, based on quantification of plastic rearrangements and on dissipated energy. Strikingly, the fraction of particles that undergo plastic rearrangements, and a percolation transition they undergo, are predictive of failure. We also find a robust power law relationship between accumulated damage, quantified by dissipated energy or non-affine displacements, and the failure times, which permits prediction of failure times based on behavior in the initial cycles. These observations reveal salient new microscopic features of fatigue failure and suggest approaches for developing a full microscopic picture of fatigue failure in amorphous solids.

Large language models (LLM) have demonstrated the ability to understand human language by leveraging large amount of text data. Automatic speech recognition (ASR) systems are often limited by available transcribed speech data and benefit from a second pass rescoring using LLM. Recently multi-modal large language models, particularly speech and text foundational models have demonstrated strong spoken language understanding. Speech-Text foundational models leverage large amounts of unlabelled and labelled data both in speech and text modalities to model human language. In this work, we propose novel techniques to use multi-modal LLM for ASR rescoring. We also explore discriminative training to further improve the foundational model rescoring performance. We demonstrate cross-modal knowledge transfer in speech-text LLM can benefit rescoring. Our experiments demonstrate up-to 20% relative improvements over Whisper large ASR and up-to 15% relative improvements over text-only LLM.

Digital Twins (DTs) are becoming popular in Additive Manufacturing (AM) due to their ability to create virtual replicas of physical components of AM machines, which helps in real-time production monitoring. Advanced techniques such as Machine Learning (ML), Augmented Reality (AR), and simulation-based models play key roles in developing intelligent and adaptable DTs in manufacturing processes. However, questions remain regarding scalability, the integration of high-quality data, and the computational power required for real-time applications in developing DTs. Understanding the current state of DTs in AM is essential to address these challenges and fully utilize their potential in advancing AM processes. Considering this opportunity, this work aims to provide a comprehensive overview of DTs in AM by addressing the following four research questions: (1) What are the key types of DTs used in AM and their specific applications? (2) What are the recent developments and implementations of DTs? (3) How are DTs employed in process improvement and hybrid manufacturing? (4) How are DTs integrated with Industry 4.0 technologies? By discussing current applications and techniques, we aim to offer a better understanding and potential future research directions for researchers and practitioners in AM and DTs.

Using particle-based numerical simulations performed under pressure-imposed conditions, we investigate the transient dilation dynamics of a shear thickening suspension brought to shear jamming. We show that the stress levels, instead of diverging as predicted by steady state flow rules, remain finite and are entirely determined by the coupling between the particle network dilation and the resulting Darcy backflow. System-spanning stress gradients along the dilation direction lead to cross-system stress differences scaling quadratically with the system size. Measured stress levels are quantitatively captured by a continuum model based on a Reynolds-like dilatancy law and the Wyart-Cates constitutive model. Beyond globally jammed suspensions, our results enable the modeling of inhomogeneous flows where shear jamming is local, e.g. under impact, which eludes usual shear thickening rheological laws., Comment: 6 pages + 3 pages of supplementary

We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.

Diffusion Models are probabilistic models that create realistic samples by simulating the diffusion process, gradually adding and removing noise from data. These models have gained popularity in domains such as image processing, speech synthesis, and natural language processing due to their ability to produce high-quality samples. As Diffusion Models are being adopted in various domains, existing literature reviews that often focus on specific areas like computer vision or medical imaging may not serve a broader audience across multiple fields. Therefore, this review presents a comprehensive overview of Diffusion Models, covering their theoretical foundations and algorithmic innovations. We highlight their applications in diverse areas such as media quality, authenticity, synthesis, image transformation, healthcare, and more. By consolidating current knowledge and identifying emerging trends, this review aims to facilitate a deeper understanding and broader adoption of Diffusion Models and provide guidelines for future researchers and practitioners across diverse disciplines.

The development of conversational AI assistants is an iterative process with multiple components. As such, the evaluation and continual improvement of these assistants is a complex and multifaceted problem. This paper introduces the challenges in evaluating and improving a generative AI assistant for enterprises, which is under active development, and how we address these challenges. We also share preliminary results and discuss lessons learned., Comment: Accepted to DaSH Workshop at NAACL 2024

Retrieval is a widely adopted approach for improving language models leveraging external information. As the field moves towards multi-modal large language models, it is important to extend the pure text based methods to incorporate other modalities in retrieval as well for applications across the wide spectrum of machine learning tasks and data types. In this work, we propose multi-modal retrieval with two approaches: kNN-LM and cross-attention techniques. We demonstrate the effectiveness of our retrieval approaches empirically by applying them to automatic speech recognition tasks with access to external information. Under this setting, we show that speech-based multi-modal retrieval outperforms text based retrieval, and yields up to 50 % improvement in word error rate over the multi-modal language model baseline. Furthermore, we achieve state-of-the-art recognition results on the Spoken-Squad question answering dataset.

We present the Keras domain packages KerasCV and KerasNLP, extensions of the Keras API for Computer Vision and Natural Language Processing workflows, capable of running on either JAX, TensorFlow, or PyTorch. These domain packages are designed to enable fast experimentation, with a focus on ease-of-use and performance. We adopt a modular, layered design: at the library's lowest level of abstraction, we provide building blocks for creating models and data preprocessing pipelines, and at the library's highest level of abstraction, we provide pretrained ``task" models for popular architectures such as Stable Diffusion, YOLOv8, GPT2, BERT, Mistral, CLIP, Gemma, T5, etc. Task models have built-in preprocessing, pretrained weights, and can be fine-tuned on raw inputs. To enable efficient training, we support XLA compilation for all models, and run all preprocessing via a compiled graph of TensorFlow operations using the tf.data API. The libraries are fully open-source (Apache 2.0 license) and available on GitHub., Comment: Submitted to Journal of Machine Learning Open Source Software

Data privacy has become a major concern in healthcare due to the increasing digitization of medical records and data-driven medical research. Protecting sensitive patient information from breaches and unauthorized access is critical, as such incidents can have severe legal and ethical complications. Federated Learning (FL) addresses this concern by enabling multiple healthcare institutions to collaboratively learn from decentralized data without sharing it. FL's scope in healthcare covers areas such as disease prediction, treatment customization, and clinical trial research. However, implementing FL poses challenges, including model convergence in non-IID (independent and identically distributed) data environments, communication overhead, and managing multi-institutional collaborations. A systematic review of FL in healthcare is necessary to evaluate how effectively FL can provide privacy while maintaining the integrity and usability of medical data analysis. In this study, we analyze existing literature on FL applications in healthcare. We explore the current state of model security practices, identify prevalent challenges, and discuss practical applications and their implications. Additionally, the review highlights promising future research directions to refine FL implementations, enhance data security protocols, and expand FL's use to broader healthcare applications, which will benefit future researchers and practitioners.

Data race, a category of insidious software concurrency bugs, is often challenging and resource-intensive to detect and debug. Existing dynamic race detection tools incur significant execution time and memory overhead while exhibiting high false positives. This paper proposes HMTRace, a novel Armv8.5-A memory tag extension (MTE) based dynamic data race detection framework, emphasizing low compute and memory requirements while maintaining high accuracy and precision. HMTRace supports race detection in userspace OpenMP- and Pthread-based multi-threaded C applications. HMTRace showcases a combined f1-score of 0.86 while incurring a mean execution time overhead of 4.01% and peak memory (RSS) overhead of 54.31%. HMTRace also does not report false positives, asserting all reported races.

Additive manufacturing (AM), particularly 3D printing, has revolutionized the production of complex structures across various industries. However, ensuring quality and detecting defects in 3D-printed objects remain significant challenges. This study focuses on improving defect detection in 3D-printed cylinders by integrating novel pre-processing techniques such as Region of Interest (ROI) selection, Histogram Equalization (HE), and Details Enhancer (DE) with Convolutional Neural Networks (CNNs), specifically the modified VGG16 model. The approaches, ROIN, ROIHEN, and ROIHEDEN, demonstrated promising results, with the best model achieving an accuracy of 1.00 and an F1-score of 1.00 on the test set. The study also explored the models' interpretability through Local Interpretable Model-Agnostic Explanations and Gradient-weighted Class Activation Mapping, enhancing the understanding of the decision-making process. Furthermore, the modified VGG16 model showed superior computational efficiency with 30713M FLOPs and 15M parameters, the lowest among the compared models. These findings underscore the significance of tailored pre-processing and CNNs in enhancing defect detection in AM, offering a pathway to improve manufacturing precision and efficiency. This research not only contributes to the advancement of 3D printing technology but also highlights the potential of integrating machine learning with AM for superior quality control.

This work explores the zero-shot capabilities of foundation models in Visual Question Answering (VQA) tasks. We propose an adaptive multi-agent system, named Multi-Agent VQA, to overcome the limitations of foundation models in object detection and counting by using specialized agents as tools. Unlike existing approaches, our study focuses on the system's performance without fine-tuning it on specific VQA datasets, making it more practical and robust in the open world. We present preliminary experimental results under zero-shot scenarios and highlight some failure cases, offering new directions for future research., Comment: A full version of the paper will be released soon. The codes are available at https://github.com/bowen-upenn/Multi-Agent-VQA

This paper presents SNOW-SCA, the first power side-channel analysis (SCA) attack of a 5G mobile communication security standard candidate, SNOW-V, running on a 32-bit ARM Cortex-M4 microcontroller. First, we perform a generic known-key correlation (KKC) analysis to identify the leakage points. Next, a correlation power analysis (CPA) attack is performed, which reduces the attack complexity to two key guesses for each key byte. The correct secret key is then uniquely identified utilizing linear discriminant analysis (LDA). The profiled SCA attack with LDA achieves 100% accuracy after training with $<200$ traces, which means the attack succeeds with just a single trace. Overall, using the \textit{combined CPA and LDA attack} model, the correct secret key byte is recovered with <50 traces collected using the ChipWhisperer platform. The entire 256-bit secret key of SNOW-V can be recovered incrementally using the proposed SCA attack. Finally, we suggest low-overhead countermeasures that can be used to prevent these SCA attacks.

In this paper, we present solvable, convex formulations of $H_2$-optimal state estimation and state-feedback control problems for a general class of linear Partial Differential Equations (PDEs) with one spatial dimension. These convex formulations are derived by using an analysis and control framework called the `Partial Integral Equation' (PIE) framework, which utilizes the PIE representation of infinite-dimensional systems. Since PIEs are parameterized by Partial Integral (PI) operators that form an algebra, $H_2$-optimal estimation and control problems for PIEs can be formulated as Linear PI Inequalities (LPIs). Furthermore, if a PDE admits a PIE representation, then the stability and $H_2$ performance of the PIE system implies that of the PDE system. Consequently, the optimal estimator and controller obtained for a PIE using LPIs provide the same stability and performance when applied to the corresponding PDE. These LPI optimization problems can be solved computationally using semi-definite programming solvers because such problems can be formulated using Linear Matrix Inequalities by using positive matrices to parameterize a cone of positive PI operators. We illustrate the application of these methods by constructing observers and controllers for some standard PDE examples.

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.