Search

Online learning methods, like the seminal Passive-Aggressive (PA) classifier, are still highly effective for high-dimensional streaming data, out-of-core processing, and other throughput-sensitive applications. Many such algorithms rely on fast adaptation to individual errors as a key to their convergence. While such algorithms enjoy low theoretical regret, in real-world deployment they can be sensitive to individual outliers that cause the algorithm to over-correct. When such outliers occur at the end of the data stream, this can cause the final solution to have unexpectedly low accuracy. We design a weighted reservoir sampling (WRS) approach to obtain a stable ensemble model from the sequence of solutions without requiring additional passes over the data, hold-out sets, or a growing amount of memory. Our key insight is that good solutions tend to be error-free for more iterations than bad solutions, and thus, the number of passive rounds provides an estimate of a solution's relative quality. Our reservoir thus contains $K$ previous intermediate weight vectors with high survival times. We demonstrate our WRS approach on the Passive-Aggressive Classifier (PAC) and First-Order Sparse Online Learning (FSOL), where our method consistently and significantly outperforms the unmodified approach. We show that the risk of the ensemble classifier is bounded with respect to the regret of the underlying online learning method., Comment: To appear in the 38th Conference on Neural Information Processing Systems (NeurIPS 2024)

Binary analysis is a core component of many critical security tasks, including reverse engineering, malware analysis, and vulnerability detection. Manual analysis is often time-consuming, but identifying commonly-used or previously-seen functions can reduce the time it takes to understand a new file. However, given the complexity of assembly, and the NP-hard nature of determining function equivalence, this task is extremely difficult. Common approaches often use sophisticated disassembly and decompilation tools, graph analysis, and other expensive pre-processing steps to perform function similarity searches over some corpus. In this work, we identify a number of discrepancies between the current research environment and the underlying application need. To remedy this, we build a new benchmark, REFuSE-Bench, for binary function similarity detection consisting of high-quality datasets and tests that better reflect real-world use cases. In doing so, we address issues like data duplication and accurate labeling, experiment with real malware, and perform the first serious evaluation of ML binary function similarity models on Windows data. Our benchmark reveals that a new, simple basline, one which looks at only the raw bytes of a function, and requires no disassembly or other pre-processing, is able to achieve state-of-the-art performance in multiple settings. Our findings challenge conventional assumptions that complex models with highly-engineered features are being used to their full potential, and demonstrate that simpler approaches can provide significant value., Comment: To appear in the 38th Conference on Neural Information Processing Systems (NeurIPS 2024) Track on Datasets and Benchmarks

Vector Symbolic Architectures (VSAs) are one approach to developing Neuro-symbolic AI, where two vectors in $\mathbb{R}^d$ are `bound' together to produce a new vector in the same space. VSAs support the commutativity and associativity of this binding operation, along with an inverse operation, allowing one to construct symbolic-style manipulations over real-valued vectors. Most VSAs were developed before deep learning and automatic differentiation became popular and instead focused on efficacy in hand-designed systems. In this work, we introduce the Hadamard-derived linear Binding (HLB), which is designed to have favorable computational efficiency, and efficacy in classic VSA tasks, and perform well in differentiable systems. Code is available at https://github.com/FutureComputing4AI/Hadamard-derived-Linear-Binding, Comment: To appear in the 38th Conference on Neural Information Processing Systems (NeurIPS 2024)

In this article, we seek to elucidate challenges and opportunities for differential privacy within the federal government setting, as seen by a team of differential privacy researchers, privacy lawyers, and data scientists working closely with the U.S. government. After introducing differential privacy, we highlight three significant challenges which currently restrict the use of differential privacy in the U.S. government. We then provide two examples where differential privacy can enhance the capabilities of government agencies. The first example highlights how the quantitative nature of differential privacy allows policy security officers to release multiple versions of analyses with different levels of privacy. The second example, which we believe is a novel realization, indicates that differential privacy can be used to improve staffing efficiency in classified applications. We hope that this article can serve as a nontechnical resource which can help frame future action from the differential privacy community, privacy regulators, security officers, and lawmakers., Comment: 2nd Workshop on Regulatable ML at NeurIPS 2024

It is generally well understood that predictive classification and compression are intrinsically related concepts in information theory. Indeed, many deep learning methods are explained as learning a kind of compression, and that better compression leads to better performance. We interrogate this hypothesis via the Normalized Compression Distance (NCD), which explicitly relies on compression as the means of measuring similarity between sequences and thus enables nearest-neighbor classification. By turning popular large language models (LLMs) into lossless compressors, we develop a Neural NCD and compare LLMs to classic general-purpose algorithms like gzip. In doing so, we find that classification accuracy is not predictable by compression rate alone, among other empirical aberrations not predicted by current understanding. Our results imply that our intuition on what it means for a neural network to ``compress'' and what is needed for effective classification are not yet well understood., Comment: Accepted to Machine Learning and Compression Workshop at 38th Conference on Neural Information Processing Systems

Trajectory optimization is an essential tool for generating efficient and dynamically consistent gaits in legged locomotion. This paper explores the indirect method of trajectory optimization, emphasizing its application in creating optimal periodic gaits for legged systems and contrasting it with the more commonly used direct method. While the direct method provides considerable flexibility in its implementation, it is limited by its input space parameterization. In contrast, the indirect method improves accuracy by defining control inputs as functions of the system's states and costates. We tackle the convergence challenges associated with indirect shooting methods, particularly through the systematic development of gait libraries by utilizing numerical continuation methods. Our contributions include: (1) the formalization of a general periodic trajectory optimization problem that extends existing first-order necessary conditions for a broader range of cost functions and operating conditions; (2) a methodology for efficiently generating libraries of optimal trajectories (gaits) utilizing a single shooting approach combined with numerical continuation methods, including a novel approach for reconstructing Lagrange multipliers and costates from passive gaits; and (3) a comparative analysis of the indirect and direct shooting methods using a compass-gait walker as a case study, demonstrating the former's superior accuracy in generating optimal gaits. The findings underscore the potential of the indirect method for generating families of optimal gaits, thereby advancing the field of trajectory optimization in legged robotics., Comment: submitted to the International Journal of Robotics Research (IJRR)

Data-based methods have gained increasing importance in engineering, especially but not only driven by successes with deep artificial neural networks. Success stories are prevalent, e.g., in areas such as data-driven modeling, control and automation, as well as surrogate modeling for accelerated simulation. Beyond engineering, generative and large-language models are increasingly performing and helping with tasks that, previously, were solely associated with creative human processes. Thus, it seems timely to seek artificial-intelligence-support for engineering design tasks to automate, help with, or accelerate purpose-built designs of engineering systems, e.g., in mechanics and dynamics, where design so far requires a lot of specialized knowledge. However, research-wise, compared to established, predominantly first-principles-based methods, the datasets used for training, validation, and test become an almost inherent part of the overall methodology. Thus, data publishing becomes just as important in (data-driven) engineering science as appropriate descriptions of conventional methodology in publications in the past. This article analyzes the value and challenges of data publishing in mechanics and dynamics, in particular regarding engineering design tasks, showing that the latter raise also challenges and considerations not typical in fields where data-driven methods have been booming originally. Possible ways to deal with these challenges are discussed and a set of examples from across different design problems shows how data publishing can be put into practice. The analysis, discussions, and examples are based on the research experience made in a priority program of the German research foundation focusing on research on artificially intelligent design assistants in mechanics and dynamics., Comment: 21 pages, 8 figures

An obstetric goal for a laboring mother is to achieve a vaginal delivery as it reduces the risks inherent in major abdominal surgery (i.e., a Cesarean section). Various medical interventions may be used by a physician to increase the likelihood of this occurring while minimizing maternal and fetal morbidity. However, patients with prelabor rupture of membranes (PROM) have only two commonly used options for cervical ripening, Pitocin and misoprostol. Little research exists on the benefits/risks for these two key drugs for PROM patients. A major limitation with most induction-of-labor related research is the inability to account for differences in \textit{Bishop scores} that are commonly used in obstetrical practice to determine the next induction agent offered to the patient. This creates a confounding factor, which biases the results, but has not been realized in the literature. In this work, we use a Bayesian model of the relationships between the relevant factors, informed by expert physicians, to separate the confounding variable from its actual impact. In doing so, we provide strong evidence that pitocin and buccal misoprostol are equally effective and safe; thus, physicians have more choice in clinical care than previously realized. This is particularly important for developing countries where neither medication may be readily available, and prior guidelines may create an artificial barrier to needed medication., Comment: To appear in the 2024 IEEE 11th International Conference on Data Science and Advanced Analytics (DSAA)

Data scientists often seek to identify the most important features in high-dimensional datasets. This can be done through $L_1$-regularized regression, but this can become inefficient for very high-dimensional datasets. Additionally, high-dimensional regression can leak information about individual datapoints in a dataset. In this paper, we empirically evaluate the established baseline method for feature selection with differential privacy, the two-stage selection technique, and show that it is not stable under sparsity. This makes it perform poorly on real-world datasets, so we consider a different approach to private feature selection. We employ a correlations-based order statistic to choose important features from a dataset and privatize them to ensure that the results do not leak information about individual datapoints. We find that our method significantly outperforms the established baseline for private feature selection on many datasets., Comment: To appear in Proceedings of the 17th ACM Workshop on Artificial Intelligence and Security, 2024

Previous research on testing the vulnerabilities in Large Language Models (LLMs) using adversarial attacks has primarily focused on nonsensical prompt injections, which are easily detected upon manual or automated review (e.g., via byte entropy). However, the exploration of innocuous human-understandable malicious prompts augmented with adversarial injections remains limited. In this research, we explore converting a nonsensical suffix attack into a sensible prompt via a situation-driven contextual re-writing. This allows us to show suffix conversion without any gradients, using only LLMs to perform the attacks, and thus better understand the scope of possible risks. We combine an independent, meaningful adversarial insertion and situations derived from movies to check if this can trick an LLM. The situations are extracted from the IMDB dataset, and prompts are defined following a few-shot chain-of-thought prompting. Our approach demonstrates that a successful situation-driven attack can be executed on both open-source and proprietary LLMs. We find that across many LLMs, as few as 1 attempt produces an attack and that these attacks transfer between LLMs.

As the size of datasets used in statistical learning continues to grow, distributed training of models has attracted increasing attention. These methods partition the data and exploit parallelism to reduce memory and runtime, but suffer increasingly from communication costs as the data size or the number of iterations grows. Recent work on linear models has shown that a surrogate likelihood can be optimized locally to iteratively improve on an initial solution in a communication-efficient manner. However, existing versions of these methods experience multiple shortcomings as the data size becomes massive, including diverging updates and efficiently handling sparsity. In this work we develop solutions to these problems which enable us to learn a communication-efficient distributed logistic regression model even beyond millions of features. In our experiments we demonstrate a large improvement in accuracy over distributed algorithms with only a few distributed update steps needed, and similar or faster runtimes. Our code is available at \url{https://github.com/FutureComputing4AI/ProxCSL}., Comment: KDD 2024, Research Track

Language Models (LMs) are being proposed for mental health applications where the heightened risk of adverse outcomes means predictive performance may not be a sufficient litmus test of a model's utility in clinical practice. A model that can be trusted for practice should have a correspondence between explanation and clinical determination, yet no prior research has examined the attention fidelity of these models and their effect on ground truth explanations. We introduce an evaluation design that focuses on the robustness and explainability of LMs in identifying Wellness Dimensions (WDs). We focus on two existing mental health and well-being datasets: (a) Multi-label Classification-based MultiWD, and (b) WellXplain for evaluating attention mechanism veracity against expert-labeled explanations. The labels are based on Halbert Dunn's theory of wellness, which gives grounding to our evaluation. We reveal four surprising results about LMs/LLMs: (1) Despite their human-like capabilities, GPT-3.5/4 lag behind RoBERTa, and MedAlpaca, a fine-tuned LLM on WellXplain fails to deliver any remarkable improvements in performance or explanations. (2) Re-examining LMs' predictions based on a confidence-oriented loss function reveals a significant performance drop. (3) Across all LMs/LLMs, the alignment between attention and explanations remains low, with LLMs scoring a dismal 0.0. (4) Most mental health-specific LMs/LLMs overlook domain-specific knowledge and undervalue explanations, causing these discrepancies. This study highlights the need for further research into their consistency and explanations in mental health and well-being., Comment: Accepted in BlackboxNLP @ EMNLP 2024

While distributed training is often viewed as a solution to optimizing linear models on increasingly large datasets, inter-machine communication costs of popular distributed approaches can dominate as data dimensionality increases. Recent work on non-interactive algorithms shows that approximate solutions for linear models can be obtained efficiently with only a single round of communication among machines. However, this approximation often degenerates as the number of machines increases. In this paper, building on the recent optimal weighted average method, we introduce a new technique, ACOWA, that allows an extra round of communication to achieve noticeably better approximation quality with minor runtime increases. Results show that for sparse distributed logistic regression, ACOWA obtains solutions that are more faithful to the empirical risk minimizer and attain substantially higher accuracy than other distributed algorithms., Comment: Under review

Binary code is pervasive, and binary analysis is a key task in reverse engineering, malware classification, and vulnerability discovery. Unfortunately, while there exist large corpora of malicious binaries, obtaining high-quality corpora of benign binaries for modern systems has proven challenging (e.g., due to licensing issues). Consequently, machine learning based pipelines for binary analysis utilize either costly commercial corpora (e.g., VirusTotal) or open-source binaries (e.g., coreutils) available in limited quantities. To address these issues, we present Assemblage: an extensible cloud-based distributed system that crawls, configures, and builds Windows PE binaries to obtain high-quality binary corpuses suitable for training state-of-the-art models in binary analysis. We have run Assemblage on AWS over the past year, producing 890k Windows PE and 428k Linux ELF binaries across 29 configurations. Assemblage is designed to be both reproducible and extensible, enabling users to publish "recipes" for their datasets, and facilitating the extraction of a wide array of features. We evaluated Assemblage by using its data to train modern learning-based pipelines for compiler provenance and binary function similarity. Our results illustrate the practical need for robust corpora of high-quality Windows PE binaries in training modern learning-based binary analyses. Assemblage code is open sourced under the MIT license, and the dataset can be downloaded from https://assemblage-dataset.net, Comment: To appear in the 38th Conference on Neural Information Processing Systems (NeurIPS 2024) Track on Datasets and Benchmarks

Automatic citation generation for sentences in a document or report is paramount for intelligence analysts, cybersecurity, news agencies, and education personnel. In this research, we investigate whether large language models (LLMs) are capable of generating references based on two forms of sentence queries: (a) Direct Queries, LLMs are asked to provide author names of the given research article, and (b) Indirect Queries, LLMs are asked to provide the title of a mentioned article when given a sentence from a different article. To demonstrate where LLM stands in this task, we introduce a large dataset called REASONS comprising abstracts of the 12 most popular domains of scientific research on arXiv. From around 20K research articles, we make the following deductions on public and proprietary LLMs: (a) State-of-the-art, often called anthropomorphic GPT-4 and GPT-3.5, suffers from high pass percentage (PP) to minimize the hallucination rate (HR). When tested with Perplexity.ai (7B), they unexpectedly made more errors; (b) Augmenting relevant metadata lowered the PP and gave the lowest HR; (c) Advance retrieval-augmented generation (RAG) using Mistral demonstrates consistent and robust citation support on indirect queries and matched performance to GPT-3.5 and GPT-4. The HR across all domains and models decreased by an average of 41.93%, and the PP was reduced to 0% in most cases. In terms of generation quality, the average F1 Score and BLEU were 68.09% and 57.51%, respectively; (d) Testing with adversarial samples showed that LLMs, including the Advance RAG Mistral, struggle to understand context, but the extent of this issue was small in Mistral and GPT-4-Preview. Our study contributes valuable insights into the reliability of RAG for automated citation generation tasks., Comment: Work in progress

We identify the nonlinear normal modes spawning from the stable equilibrium of a double pendulum under gravity, and we establish their connection to homoclinic orbits through the unstable upright position as energy increases. This result is exploited to devise an efficient swing-up strategy for a double pendulum with weak, saturating actuators. Our approach involves stabilizing the system onto periodic orbits associated with the nonlinear modes while gradually injecting energy. Since these modes are autonomous system evolutions, the required control effort for stabilization is minimal. Even with actuator limitations of less than 1% of the maximum gravitational torque, the proposed method accomplishes the swing-up of the double pendulum by allowing sufficient time., Comment: Preprint of a paper to appear at the European Control Conference (ECC) 2024 in Stockholm, Sweden

Linear models are ubiquitous in data science, but are particularly prone to overfitting and data memorization in high dimensions. To guarantee the privacy of training data, differential privacy can be used. Many papers have proposed optimization techniques for high-dimensional differentially private linear models, but a systematic comparison between these methods does not exist. We close this gap by providing a comprehensive review of optimization methods for private high-dimensional linear models. Empirical tests on all methods demonstrate robust and coordinate-optimized algorithms perform best, which can inform future research. Code for implementing all methods is released online., Comment: 21 pages, 7 figures. To be published at the 2nd IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)

Malware detection is an interesting and valuable domain to work in because it has significant real-world impact and unique machine-learning challenges. We investigate existing long-range techniques and benchmarks and find that they're not very suitable in this problem area. In this paper, we introduce Holographic Global Convolutional Networks (HGConv) that utilize the properties of Holographic Reduced Representations (HRR) to encode and decode features from sequence elements. Unlike other global convolutional methods, our method does not require any intricate kernel computation or crafted kernel design. HGConv kernels are defined as simple parameters learned through backpropagation. The proposed method has achieved new SOTA results on Microsoft Malware Classification Challenge, Drebin, and EMBER malware benchmarks. With log-linear complexity in sequence length, the empirical results demonstrate substantially faster run-time by HGConv compared to other methods achieving far more efficient scaling even with sequence length $\geq 100,000$., Comment: To appear in Proceedings of the 27th International Conference on Artificial Intelligence and Statistics (AISTATS) 2024, Valencia, Spain

In autonomous differential equations where a single first integral is present, periodic orbits are well-known to belong to one-parameter families, parameterized by the first integral's values. This paper shows that this characteristic extends to a broader class of conservative hybrid dynamical systems (cHDSs). We study periodic orbits of a cHDS, introducing the concept of a hybrid first integral to characterize conservation in these systems. Additionally, our work presents a methodology that utilizes numerical continuation methods to generate these periodic orbits, building upon the concept of normal periodic orbits introduced by Sepulchre and MacKay (1997). We specifically compare state-based and time-based implementations of an cHDS as an important application detail in generating periodic orbits. Furthermore, we showcase the continuation process using exemplary conservative mechanical systems with impulsive dynamics., Comment: accepted for publication in Nonlinear Dynamics (October 25, 2024)

Industry practitioners care about small improvements in malware detection accuracy because their models are deployed to hundreds of millions of machines, meaning a 0.1\% change can cause an overwhelming number of false positives. However, academic research is often restrained to public datasets on the order of ten thousand samples and is too small to detect improvements that may be relevant to industry. Working within these constraints, we devise an approach to generate a benchmark of configurable difficulty from a pool of available samples. This is done by leveraging malware family information from tools like AVClass to construct training/test splits that have different generalization rates, as measured by a secondary model. Our experiments will demonstrate that using a less accurate secondary model with disparate features is effective at producing benchmarks for a more sophisticated target model that is under evaluation. We also ablate against alternative designs to show the need for our approach., Comment: To appear in Conference on Applied Machine Learning for Information Security 2023

While deep learning has enjoyed significant success in computer vision tasks over the past decade, many shortcomings still exist from a Cognitive Science (CogSci) perspective. In particular, the ability to subitize, i.e., quickly and accurately identify the small (less than 6) count of items, is not well learned by current Convolutional Neural Networks (CNNs) or Vision Transformers (ViTs) when using a standard cross-entropy (CE) loss. In this paper, we demonstrate that adapting tools used in CogSci research can improve the subitizing generalization of CNNs and ViTs by developing an alternative loss function using Holographic Reduced Representations (HRRs). We investigate how this neuro-symbolic approach to learning affects the subitizing capability of CNNs and ViTs, and so we focus on specially crafted problems that isolate generalization to specific aspects of subitizing. Via saliency maps and out-of-distribution performance, we are able to empirically observe that the proposed HRR loss improves subitizing generalization though it does not completely solve the problem. In addition, we find that ViTs perform considerably worse compared to CNNs in most respects on subitizing, except on one axis where an HRR-based loss provides improvement., Comment: Accepted in 38th Annual AAAI Workshop on Neuro-Symbolic Learning and Reasoning in the Era of Large Language Models (NuCLeaR), 2024

Differential Diagnosis (DDx) is the process of identifying the most likely medical condition among the possible pathologies through the process of elimination based on evidence. An automated process that narrows a large set of pathologies down to the most likely pathologies will be of great importance. The primary prior works have relied on the Reinforcement Learning (RL) paradigm under the intuition that it aligns better with how physicians perform DDx. In this paper, we show that a generative approach trained with simpler supervised and self-supervised learning signals can achieve superior results on the current benchmark. The proposed Transformer-based generative network, named DDxT, autoregressively produces a set of possible pathologies, i.e., DDx, and predicts the actual pathology using a neural network. Experiments are performed using the DDXPlus dataset. In the case of DDx, the proposed network has achieved a mean accuracy of 99.82% and a mean F1 score of 0.9472. Additionally, mean accuracy reaches 99.98% with a mean F1 score of 0.9949 while predicting ground truth pathology. The proposed DDxT outperformed the previous RL-based approaches by a big margin. Overall, the automated Transformer-based DDx generative model has the potential to become a useful tool for a physician in times of urgency., Comment: Accepted at 1st Workshop on Deep Generative Models for Health at NeurIPS 2023

To the best of our knowledge, there are no methods today for training differentially private regression models on sparse input data. To remedy this, we adapt the Frank-Wolfe algorithm for $L_1$ penalized linear regression to be aware of sparse inputs and to use them effectively. In doing so, we reduce the training time of the algorithm from $\mathcal{O}( T D S + T N S)$ to $\mathcal{O}(N S + T \sqrt{D} \log{D} + T S^2)$, where $T$ is the number of iterations and a sparsity rate $S$ of a dataset with $N$ rows and $D$ features. Our results demonstrate that this procedure can reduce runtime by a factor of up to $2,200\times$, depending on the value of the privacy parameter $\epsilon$ and the sparsity of the dataset., Comment: To appear in the 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

Multiple Instance Learning (MIL) is a sub-domain of classification problems with positive and negative labels and a "bag" of inputs, where the label is positive if and only if a positive element is contained within the bag, and otherwise is negative. Training in this context requires associating the bag-wide label to instance-level information, and implicitly contains a causal assumption and asymmetry to the task (i.e., you can't swap the labels without changing the semantics). MIL problems occur in healthcare (one malignant cell indicates cancer), cyber security (one malicious executable makes an infected computer), and many other tasks. In this work, we examine five of the most prominent deep-MIL models and find that none of them respects the standard MIL assumption. They are able to learn anti-correlated instances, i.e., defaulting to "positive" labels until seeing a negative counter-example, which should not be possible for a correct MIL model. We suspect that enhancements and other works derived from these models will share the same issue. In any context in which these models are being used, this creates the potential for learning incorrect models, which creates risk of operational failure. We identify and demonstrate this problem via a proposed "algorithmic unit test", where we create synthetic datasets that can be solved by a MIL respecting model, and which clearly reveal learning that violates MIL assumptions. The five evaluated methods each fail one or more of these tests. This provides a model-agnostic way to identify violations of modeling assumptions, which we hope will be useful for future development and evaluation of MIL models., Comment: To appear in the 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

Existing research on malware classification focuses almost exclusively on two tasks: distinguishing between malicious and benign files and classifying malware by family. However, malware can be categorized according to many other types of attributes, and the ability to identify these attributes in newly-emerging malware using machine learning could provide significant value to analysts. In particular, we have identified four tasks which are under-represented in prior work: classification by behaviors that malware exhibit, platforms that malware run on, vulnerabilities that malware exploit, and packers that malware are packed with. To obtain labels for training and evaluating ML classifiers on these tasks, we created an antivirus (AV) tagging tool called ClarAVy. ClarAVy's sophisticated AV label parser distinguishes itself from prior AV-based taggers, with the ability to accurately parse 882 different AV label formats used by 90 different AV products. We are releasing benchmark datasets for each of these four classification tasks, tagged using ClarAVy and comprising nearly 5.5 million malicious files in total. Our malware behavior dataset includes 75 distinct tags - nearly 7x more than the only prior benchmark dataset with behavioral tags. To our knowledge, we are the first to release datasets with malware platform and packer tags.

Intuitively, students who start a homework assignment earlier and spend more time on it should receive better grades on the assignment. However, existing literature on the impact of time spent on homework is not clear-cut and comes mostly from K-12 education. It is not clear that these prior studies can inform coursework in deep learning due to differences in demographics, as well as the computational time needed for assignments to be completed. We study this problem in a post-hoc study of three semesters of a deep learning course at the University of Maryland, Baltimore County (UMBC), and develop a hierarchical Bayesian model to help make principled conclusions about the impact on student success given an approximate measure of the total time spent on the homework, and how early they submitted the assignment. Our results show that both submitting early and spending more time positively relate with final grade. Surprisingly, the value of an additional day of work is apparently equal across students, even when some require less total time to complete an assignment., Comment: To appear in AI for AI Education, co-located with ECAI 2023

Rejection sampling is a common tool for low dimensional problems ($d \leq 2$), often touted as an "easy" way to obtain valid samples from a distribution $f(\cdot)$ of interest. In practice it is non-trivial to apply, often requiring considerable mathematical effort to devise a good proposal distribution $g(\cdot)$ and select a supremum $C$. More advanced samplers require additional mathematical derivations, limitations on $f(\cdot)$, or even cross-validation, making them difficult to apply. We devise a new approximate baseline approach to rejection sampling that works with less information, requiring only a differentiable $f(\cdot)$ be specified, making it easier to use. We propose a new approach to rejection sampling by refining a parameterized proposal distribution with a loss derived from the acceptance threshold. In this manner we obtain comparable or better acceptance rates on current benchmarks by up to $7.3\times$, while requiring no extra assumptions or any derivations to use: only a differentiable $f(\cdot)$ is required. While approximate, the results are correct with high probability, and in all tests pass a distributional check. This makes our approach easy to use, reproduce, and efficacious., Comment: To appear in the 26th European Conference on Artificial Intelligence ECAI 2023

Identification of the family to which a malware specimen belongs is essential in understanding the behavior of the malware and developing mitigation strategies. Solutions proposed by prior work, however, are often not practicable due to the lack of realistic evaluation factors. These factors include learning under class imbalance, the ability to identify new malware, and the cost of production-quality labeled data. In practice, deployed models face prominent, rare, and new malware families. At the same time, obtaining a large quantity of up-to-date labeled malware for training a model can be expensive. In this paper, we address these problems and propose a novel hierarchical semi-supervised algorithm, which we call the HNMFk Classifier, that can be used in the early stages of the malware family labeling process. Our method is based on non-negative matrix factorization with automatic model selection, that is, with an estimation of the number of clusters. With HNMFk Classifier, we exploit the hierarchical structure of the malware data together with a semi-supervised setup, which enables us to classify malware families under conditions of extreme class imbalance. Our solution can perform abstaining predictions, or rejection option, which yields promising results in the identification of novel malware families and helps with maintaining the performance of the model when a low quantity of labeled data is used. We perform bulk classification of nearly 2,900 both rare and prominent malware families, through static analysis, using nearly 388,000 samples from the EMBER-2018 corpus. In our experiments, we surpass both supervised and semi-supervised baseline models with an F1 score of 0.80., Comment: Accepted at ACM TOPS

Since thermal imagery offers a unique modality to investigate pain, the U.S. National Institutes of Health (NIH) has collected a large and diverse set of cancer patient facial thermograms for AI-based pain research. However, differing angles from camera capture between thermal and visible sensors has led to misalignment between Visible-Thermal (VT) images. We modernize the classic computer vision task of image registration by applying and modifying a generative alignment algorithm to register VT cancer faces, without the need for a reference or alignment parameters. By registering VT faces, we demonstrate that the quality of thermal images produced in the generative AI downstream task of Visible-to-Thermal (V2T) image translation significantly improves up to 52.5\%, than without registration. Images in this paper have been approved by the NIH NCI for public dissemination., Comment: 2nd Annual Artificial Intelligence over Infrared Images for Medical Applications Workshop (AIIIMA) at the 26th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2023)

Convolutional layers have long served as the primary workhorse for image classification. Recently, an alternative to convolution was proposed using the Sharpened Cosine Similarity (SCS), which in theory may serve as a better feature detector. While multiple sources report promising results, there has not been to date a full-scale empirical analysis of neural network performance using these new layers. In our work, we explore SCS's parameter behavior and potential as a drop-in replacement for convolutions in multiple CNN architectures benchmarked on CIFAR-10. We find that while SCS may not yield significant increases in accuracy, it may learn more interpretable representations. We also find that, in some circumstances, SCS may confer a slight increase in adversarial robustness., Comment: Accepted to I Can't Believe It's Not Better Workshop (ICBINB) at NeurIPS 2022

In this paper, we propose cuSLINK, a novel and state-of-the-art reformulation of the SLINK algorithm on the GPU which requires only $O(Nk)$ space and uses a parameter $k$ to trade off space and time. We also propose a set of novel and reusable building blocks that compose cuSLINK. These building blocks include highly optimized computational patterns for $k$-NN graph construction, spanning trees, and dendrogram cluster extraction. We show how we used our primitives to implement cuSLINK end-to-end on the GPU, further enabling a wide range of real-world data mining and machine learning applications that were once intractable. In addition to being a primary computational bottleneck in the popular HDBSCAN algorithm, the impact of our end-to-end cuSLINK algorithm spans a large range of important applications, including cluster analysis in social and computer networks, natural language processing, and computer vision. Users can obtain cuSLINK at https://docs.rapids.ai/api/cuml/latest/api/#agglomerative-clustering, Comment: To appear in ECML PKDD 2023 by Springer Nature

Differential privacy (DP) is the prevailing technique for protecting user data in machine learning models. However, deficits to this framework include a lack of clarity for selecting the privacy budget $\epsilon$ and a lack of quantification for the privacy leakage for a particular data row by a particular trained model. We make progress toward these limitations and a new perspective by which to visualize DP results by studying a privacy metric that quantifies the extent to which a model trained on a dataset using a DP mechanism is ``covered" by each of the distributions resulting from training on neighboring datasets. We connect this coverage metric to what has been established in the literature and use it to rank the privacy of individual samples from the training set in what we call a privacy profile. We additionally show that the privacy profile can be used to probe an observed transition to indistinguishability that takes place in the neighboring distributions as $\epsilon$ decreases, which we suggest is a tool that can enable the selection of $\epsilon$ by the ML practitioner wishing to make use of DP., Comment: Approved for Public Release; Distribution Unlimited. PA #:AFRL-2022-3639

The robustness of modern machine learning (ML) models has become an increasing concern within the community. The ability to subvert a model into making errant predictions using seemingly inconsequential changes to input is startling, as is our lack of success in building models robust to this concern. Existing research shows progress, but current mitigations come with a high cost and simultaneously reduce the model's accuracy. However, such trade-offs may not be necessary when other design choices could subvert the risk. In this survey we review the current literature on attacks and their real-world occurrences, or limited evidence thereof, to critically evaluate the real-world risks of adversarial machine learning (AML) for the average entity. This is done with an eye toward how one would then mitigate these attacks in practice, the risks for production deployment, and how those risks could be managed. In doing so we elucidate that many AML threats do not warrant the cost and trade-offs of robustness due to a low likelihood of attack or availability of superior non-ML mitigations. Our analysis also recommends cases where an actor should be concerned about AML to the degree where robust ML models are necessary for a complete deployment.

For a variety of biometric cross-spectral tasks, Visible-Thermal (VT) facial pairs are used. However, due to a lack of calibration in the lab, photographic capture between two different sensors leads to severely misaligned pairs that can lead to poor results for person re-identification and generative AI. To solve this problem, we introduce our approach for VT image registration called Vista Morph. Unlike existing VT facial registration that requires manual, hand-crafted features for pixel matching and/or a supervised thermal reference, Vista Morph is completely unsupervised without the need for a reference. By learning the affine matrix through a Vision Transformer (ViT)-based Spatial Transformer Network (STN) and Generative Adversarial Networks (GAN), Vista Morph successfully aligns facial and non-facial VT images. Our approach learns warps in Hard, No, and Low-light visual settings and is robust to geometric perturbations and erasure at test time. We conduct a downstream generative AI task to show that registering training data with Vista Morph improves subject identity of generated thermal faces when performing V2T image translation.

When investigating a malicious file, searching for related files is a common task that malware analysts must perform. Given that production malware corpora may contain over a billion files and consume petabytes of storage, many feature extraction and similarity search approaches are computationally infeasible. Our work explores the potential of antivirus (AV) scan data as a scalable source of features for malware. This is possible because AV scan reports are widely available through services such as VirusTotal and are ~100x smaller than the average malware sample. The information within an AV scan report is abundant with information and can indicate a malicious file's family, behavior, target operating system, and many other characteristics. We introduce AVScan2Vec, a language model trained to comprehend the semantics of AV scan data. AVScan2Vec ingests AV scan data for a malicious file and outputs a meaningful vector representation. AVScan2Vec vectors are ~3 to 85x smaller than popular alternatives in use today, enabling faster vector comparisons and lower memory usage. By incorporating Dynamic Continuous Indexing, we show that nearest-neighbor queries on AVScan2Vec vectors can scale to even the largest malware production datasets. We also demonstrate that AVScan2Vec vectors are superior to other leading malware feature vector representations across nearly all classification, clustering, and nearest-neighbor lookup algorithms that we evaluated.

Concept erasure aims to remove specified features from a representation. It can improve fairness (e.g. preventing a classifier from using gender or race) and interpretability (e.g. removing a concept to observe changes in model behavior). We introduce LEAst-squares Concept Erasure (LEACE), a closed-form method which provably prevents all linear classifiers from detecting a concept while changing the representation as little as possible, as measured by a broad class of norms. We apply LEACE to large language models with a novel procedure called "concept scrubbing," which erases target concept information from every layer in the network. We demonstrate our method on two tasks: measuring the reliance of language models on part-of-speech information, and reducing gender bias in BERT embeddings. Code is available at https://github.com/EleutherAI/concept-erasure.

In recent years, self-attention has become the dominant paradigm for sequence modeling in a variety of domains. However, in domains with very long sequence lengths the $\mathcal{O}(T^2)$ memory and $\mathcal{O}(T^2 H)$ compute costs can make using transformers infeasible. Motivated by problems in malware detection, where sequence lengths of $T \geq 100,000$ are a roadblock to deep learning, we re-cast self-attention using the neuro-symbolic approach of Holographic Reduced Representations (HRR). In doing so we perform the same high-level strategy of the standard self-attention: a set of queries matching against a set of keys, and returning a weighted response of the values for each key. Implemented as a ``Hrrformer'' we obtain several benefits including $\mathcal{O}(T H \log H)$ time complexity, $\mathcal{O}(T H)$ space complexity, and convergence in $10\times$ fewer epochs. Nevertheless, the Hrrformer achieves near state-of-the-art accuracy on LRA benchmarks and we are able to learn with just a single layer. Combined, these benefits make our Hrrformer the first viable Transformer for such long malware classification sequences and up to $280\times$ faster to train on the Long Range Arena benchmark. Code is available at \url{https://github.com/NeuromorphicComputationResearchProgram/Hrrformer}, Comment: To appear in Proceedings of the 40th International Conference on Machine Learning (ICML)

LASSO regularized logistic regression is particularly useful for its built-in feature selection, allowing coefficients to be removed from deployment and producing sparse solutions. Differentially private versions of LASSO logistic regression have been developed, but generally produce dense solutions, reducing the intrinsic utility of the LASSO penalty. In this paper, we present a differentially private method for sparse logistic regression that maintains hard zeros. Our key insight is to first train a non-private LASSO logistic regression model to determine an appropriate privatized number of non-zero coefficients to use in final model selection. To demonstrate our method's performance, we run experiments on synthetic and real-world datasets., Comment: 20 pages, 5 figures

Memorization, or the tendency of large language models (LLMs) to output entire sequences from their training data verbatim, is a key concern for safely deploying language models. In particular, it is vital to minimize a model's memorization of sensitive datapoints such as those containing personal identifiable information (PII). The prevalence of such undesirable memorization can pose issues for model trainers, and may even require discarding an otherwise functional model. We therefore seek to predict which sequences will be memorized before a large model's full train-time by extrapolating the memorization behavior of lower-compute trial runs. We measure memorization of the Pythia model suite and plot scaling laws for forecasting memorization, allowing us to provide equi-compute recommendations to maximize the reliability (recall) of such predictions. We additionally provide further novel discoveries on the distribution of memorization scores across models and data. We release all code and data necessary to reproduce the results in this paper at https://github.com/EleutherAI/pythia

How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \url{https://github.com/EleutherAI/pythia}., Comment: Code at https://github.com/EleutherAI/pythia

For a class of biped robots with impulsive dynamics and a non-empty set of passive gaits (unactuated, periodic motions of the biped model), we present a method for computing continuous families of locally optimal gaits with respect to a class of commonly used energetic cost functions (e.g., the integral of torque-squared). We compute these families using only the passive gaits of the biped, which are globally optimal gaits with respect to these cost functions. Our approach fills in an important gap in the literature when computing a library of locally optimal gaits, which often do not make use of these globally optimal solutions as seed values. We demonstrate our approach on a well-studied two-link biped model., Comment: Published in the IROS 2023 proceedings. This version has several updated sections and figures. The proceedings contains a supplementary video, see https://ieeexplore.ieee.org/document/10342322/media#media . An implementation of the results are also available, see https://github.com/nr-codes/OptimalGaitsForCompassGait

Linear $L_1$-regularized models have remained one of the simplest and most effective tools in data analysis, especially in information retrieval problems where n-grams over text with TF-IDF or Okapi feature values are a strong and easy baseline. Over the past decade, screening rules have risen in popularity as a way to reduce the runtime for producing the sparse regression weights of $L_1$ models. However, despite the increasing need of privacy-preserving models in information retrieval, to the best of our knoweledge, no differentially private screening rule exists. In this paper, we develop the first differentially private screening rule for linear and logistic regression. In doing so, we discover difficulties in the task of making a useful private screening rule due to the amount of noise added to ensure privacy. We provide theoretical arguments and experimental evidence that this difficulty arises from the screening step itself and not the private optimizer. Based on our results, we highlight that developing an effective private $L_1$ screening method is an open problem in the differential privacy literature., Comment: 5 pages, 2 figures