Detecting dynamic patterns of task-specific responses shared across heterogeneous datasets is an essential and challenging problem in many scientific applications in medical science and neuroscience. In our motivating example of rodent electrophysiological data, identifying the dynamical patterns in neuronal activity associated with ongoing cognitive demands and behavior is key to uncovering the neural mechanisms of memory. One of the greatest challenges in investigating a cross-subject biological process is that the systematic heterogeneity across individuals could significantly undermine the power of existing machine learning methods to identify the underlying biological dynamics. In addition, many technically challenging neurobiological experiments are conducted on only a handful of subjects where rich longitudinal data are available for each subject. The low sample sizes of such experiments could further reduce the power to detect common dynamic patterns among subjects. In this paper, we propose a novel heterogeneous data integration framework based on optimal transport to extract shared patterns in complex biological processes. The key advantages of the proposed method are that it can increase discriminating power in identifying common patterns by reducing heterogeneity unrelated to the signal by aligning the extracted latent spatiotemporal information across subjects. Our approach is effective even with a small number of subjects, and does not require auxiliary matching information for the alignment. In particular, our method can align longitudinal data across heterogeneous subjects in a common latent space to capture the dynamics of shared patterns while utilizing temporal dependency within subjects.

BLAS Level 3 operations are essential for scientific computing, but finding the optimal number of threads for multi-threaded implementations on modern multi-core systems is challenging. We present an extension to the Architecture and Data-Structure Aware Linear Algebra (ADSALA) library that uses machine learning to optimize the runtime of all BLAS Level 3 operations. Our method predicts the best number of threads for each operation based on the matrix dimensions and the system architecture. We test our method on two HPC platforms with Intel and AMD processors, using MKL and BLIS as baseline BLAS implementations. We achieve speedups of 1.5 to 3.0 for all operations, compared to using the maximum number of threads. We also analyze the runtime patterns of different BLAS operations and explain the sources of speedup. Our work shows the effectiveness and generality of the ADSALA approach for optimizing BLAS routines on modern multi-core systems.

Imagine training a machine learning model with Differentially Private Stochastic Gradient Descent (DP-SGD), only to discover post-training that the noise level was either too high, crippling your model's utility, or too low, compromising privacy. The dreaded realization hits: you must start the lengthy training process from scratch. But what if you could avoid this retraining nightmare? In this study, we introduce a groundbreaking approach (to our knowledge) that applies differential privacy noise to the model's weights after training. We offer a comprehensive mathematical proof for this novel approach's privacy bounds, use formal methods to validate its privacy guarantees, and empirically evaluate its effectiveness using membership inference attacks and performance evaluations. This method allows for a single training run, followed by post-hoc noise adjustments to achieve optimal privacy-utility trade-offs. We compare this novel fine-tuned model (DP-Weights model) to a traditional DP-SGD model, demonstrating that our approach yields statistically similar performance and privacy guarantees. Our results validate the efficacy of post-training noise application, promising significant time savings and flexibility in fine-tuning differential privacy parameters, making it a practical alternative for deploying differentially private models in real-world scenarios.

Mass spectrometry is a widely used method to study molecules and processes in medicine, life sciences, chemistry, catalysis, and industrial product quality control, among many other applications. One of the main features of some mass spectrometry techniques is the extensive level of characterization (especially when coupled with chromatography and ion mobility methods, or a part of tandem mass spectrometry experiment) and a large amount of generated data per measurement. Terabyte scales can be easily reached with mass spectrometry studies. Consequently, mass spectrometry has faced the challenge of a high level of data disappearance. Researchers often neglect and then altogether lose access to the rich information mass spectrometry experiments could provide. With the development of machine learning methods, the opportunity arises to unlock the potential of these data, enabling previously inaccessible discoveries. The present perspective highlights reevaluation of mass spectrometry data analysis in the new generation of methods and describes significant challenges in the field, particularly related to problems involving the use of electrospray ionization. We argue that further applications of machine learning raise new requirements for instrumentation (increasing throughput and information density, decreasing pricing, and making more automation-friendly software), and once met, the field may experience significant transformation.

We use transductive regression techniques to learn mappings between source and target features of given parallel corpora and use these mappings to generate machine translation outputs. We show the effectiveness of $L_1$ regularized regression (\textit{lasso}) to learn the mappings between sparsely observed feature sets versus $L_2$ regularized regression. Proper selection of training instances plays an important role to learn correct feature mappings within limited computational resources and at expected accuracy levels. We introduce \textit{dice} instance selection method for proper selection of training instances, which plays an important role to learn correct feature mappings for improving the source and target coverage of the training set. We show that $L_1$ regularized regression performs better than $L_2$ regularized regression both in regression measurements and in the translation experiments using graph decoding. We present encouraging results when translating from German to English and Spanish to English. We also demonstrate results when the phrase table of a phrase-based decoder is replaced with the mappings we find with the regression model.

A primary cost driver for training large models is wall-clock training time. We show that popular time estimates based on FLOPs are poor estimates, and construct a more accurate proxy based on memory copies. We show that with some simple accounting, we can estimate the training speed of a transformer model from its hyperparameters. Combined with a scaling law curve like Chinchilla, this lets us estimate the final loss of the model. We fit our estimate to real data with a linear regression, and apply the result to rewrite Chinchilla in terms of a model's estimated training time as opposed to the amount of training data. This gives an expression for the loss in terms of the model's hyperparameters alone. We show that this expression is accurate across a wide range of model hyperparameter values, enabling us to analytically make architectural decisions and train models more efficiently.

Deep learning has achieved tremendous success. \nj{However,} unlike SVMs, which provide direct decision criteria and can be trained with a small dataset, it still has significant weaknesses due to its requirement for massive datasets during training and the black-box characteristics on decision criteria. \nj{This paper addresses} these issues by identifying support vectors in deep learning models. To this end, we propose the DeepKKT condition, an adaptation of the traditional Karush-Kuhn-Tucker (KKT) condition for deep learning models, and confirm that generated Deep Support Vectors (DSVs) using this condition exhibit properties similar to traditional support vectors. This allows us to apply our method to few-shot dataset distillation problems and alleviate the black-box characteristics of deep learning models. Additionally, we demonstrate that the DeepKKT condition can transform conventional classification models into generative models with high fidelity, particularly as latent \jh{generative} models using class labels as latent variables. We validate the effectiveness of DSVs \nj{using common datasets (ImageNet, CIFAR10 \nj{and} CIFAR100) on the general architectures (ResNet and ConvNet)}, proving their practical applicability. (See Fig.~\ref{fig:generated})

Incremental gradient and incremental proximal methods are a fundamental class of optimization algorithms used for solving finite sum problems, broadly studied in the literature. Yet, without strong convexity, their convergence guarantees have primarily been established for the ergodic (average) iterate. Motivated by applications in continual learning, we obtain the first convergence guarantees for the last iterate of both incremental gradient and incremental proximal methods, in general convex smooth (for both) and convex Lipschitz (for the proximal variants) settings. Our oracle complexity bounds for the last iterate nearly match (i.e., match up to a square-root-log or a log factor) the best known oracle complexity bounds for the average iterate, for both classes of methods. We further obtain generalizations of our results to weighted averaging of the iterates with increasing weights and for randomly permuted ordering of updates. We study incremental proximal methods as a model of continual learning with generalization and argue that large amount of regularization is crucial to preventing catastrophic forgetting. Our results generalize last iterate guarantees for incremental methods compared to state of the art, as such results were previously known only for overparameterized linear models, which correspond to convex quadratic problems with infinitely many solutions.

Combining electroencephalogram (EEG) datasets for supervised machine learning (ML) is challenging due to session, subject, and device variability. ML algorithms typically require identical features at train and test time, complicating analysis due to varying sensor numbers and positions across datasets. Simple channel selection discards valuable data, leading to poorer performance, especially with datasets sharing few channels. To address this, we propose an unsupervised approach leveraging EEG signal physics. We map EEG channels to fixed positions using field interpolation, facilitating source-free domain adaptation. Leveraging Riemannian geometry classification pipelines and transfer learning steps, our method demonstrates robust performance in brain-computer interface (BCI) tasks and potential biomarker applications. Comparative analysis against a statistical-based approach known as Dimensionality Transcending, a signal-based imputation called ComImp, source-dependent methods, as well as common channel selection and spherical spline interpolation, was conducted with leave-one-dataset-out validation on six public BCI datasets for a right-hand/left-hand classification task. Numerical experiments show that in the presence of few shared channels in train and test, the field interpolation consistently outperforms other methods, demonstrating enhanced classification performance across all datasets. When more channels are shared, field interpolation was found to be competitive with other methods and faster to compute than source-dependent methods.

The prosperity of machine learning has also brought people's concerns about data privacy. Among them, inference attacks can implement privacy breaches in various MLaaS scenarios and model training/prediction phases. Specifically, inference attacks can perform privacy inference on undisclosed target training sets based on outputs of the target model, including but not limited to statistics, membership, semantics, data representation, etc. For instance, infer whether the target data has the characteristics of AIDS. In addition, the rapid development of the machine learning community in recent years, especially the surge of model types and application scenarios, has further stimulated the inference attacks' research. Thus, studying inference attacks and analyzing them in depth is urgent and significant. However, there is still a gap in the systematic discussion of inference attacks from taxonomy, global perspective, attack, and defense perspectives. This survey provides an in-depth and comprehensive inference of attacks and corresponding countermeasures in ML-as-a-service based on taxonomy and the latest researches. Without compromising researchers' intuition, we first propose the 3MP taxonomy based on the community research status, trying to normalize the confusing naming system of inference attacks. Also, we analyze the pros and cons of each type of inference attack, their workflow, countermeasure, and how they interact with other attacks. In the end, we point out several promising directions for researchers from a more comprehensive and novel perspective.