Machine learning studies need colossal power to process massive datasets and train neural networks to reach high accuracies, which have become gradually unsustainable. Limited by the von Neumann bottleneck, current computing architectures and methods fuel this high power consumption. Here, we present an analog computing method that harnesses chaotic nonlinear attractors to perform machine learning tasks with low power consumption. Inspired by neuromorphic computing, our model is a programmable, versatile, and generalized platform for machine learning tasks. Our mode provides exceptional performance in clustering by utilizing chaotic attractors' nonlinear mapping and sensitivity to initial conditions. When deployed as a simple analog device, it only requires milliwatt-scale power levels while being on par with current machine learning techniques. We demonstrate low errors and high accuracies with our model for regression and classification-based learning tasks.

With distributed machine learning being a prominent technique for large-scale machine learning tasks, communication complexity has become a major bottleneck for speeding up training and scaling up machine numbers. In this paper, we propose a new technique named Common randOm REconstruction(CORE), which can be used to compress the information transmitted between machines in order to reduce communication complexity without other strict conditions. Especially, our technique CORE projects the vector-valued information to a low-dimensional one through common random vectors and reconstructs the information with the same random noises after communication. We apply CORE to two distributed tasks, respectively convex optimization on linear models and generic non-convex optimization, and design new distributed algorithms, which achieve provably lower communication complexities. For example, we show for linear models CORE-based algorithm can encode the gradient vector to $\mathcal{O}(1)$-bits (against $\mathcal{O}(d)$), with the convergence rate not worse, preceding the existing results.

Active learning is a paradigm that significantly enhances the performance of machine learning models when acquiring labeled data is expensive. While several benchmarks exist for evaluating active learning strategies, their findings exhibit some misalignment. This discrepancy motivates us to develop a transparent and reproducible benchmark for the community. Our efforts result in an open-sourced implementation (https://github.com/ariapoy/active-learning-benchmark) that is reliable and extensible for future research. By conducting thorough re-benchmarking experiments, we have not only rectified misconfigurations in existing benchmark but also shed light on the under-explored issue of model compatibility, which directly causes the observed discrepancy. Resolving the discrepancy reassures that the uncertainty sampling strategy of active learning remains an effective and preferred choice for most datasets. Our experience highlights the importance of dedicating research efforts towards re-benchmarking existing benchmarks to produce more credible results and gain deeper insights.

Current adversarial attacks against speaker recognition systems (SRSs) require either white-box access or heavy black-box queries to the target SRS, thus still falling behind practical attacks against proprietary commercial APIs and voice-controlled devices. To fill this gap, we propose QFA2SR, an effective and imperceptible query-free black-box attack, by leveraging the transferability of adversarial voices. To improve transferability, we present three novel methods, tailored loss functions, SRS ensemble, and time-freq corrosion. The first one tailors loss functions to different attack scenarios. The latter two augment surrogate SRSs in two different ways. SRS ensemble combines diverse surrogate SRSs with new strategies, amenable to the unique scoring characteristics of SRSs. Time-freq corrosion augments surrogate SRSs by incorporating well-designed time-/frequency-domain modification functions, which simulate and approximate the decision boundary of the target SRS and distortions introduced during over-the-air attacks. QFA2SR boosts the targeted transferability by 20.9%-70.7% on four popular commercial APIs (Microsoft Azure, iFlytek, Jingdong, and TalentedSoft), significantly outperforming existing attacks in query-free setting, with negligible effect on the imperceptibility. QFA2SR is also highly effective when launched over the air against three wide-spread voice assistants (Google Assistant, Apple Siri, and TMall Genie) with 60%, 46%, and 70% targeted transferability, respectively.

Principal component analysis (PCA) is a widely employed statistical tool used primarily for dimensionality reduction. However, it is known to be adversely affected by the presence of outlying observations in the sample, which is quite common. Robust PCA methods using M-estimators have theoretical benefits, but their robustness drop substantially for high dimensional data. On the other end of the spectrum, robust PCA algorithms solving principal component pursuit or similar optimization problems have high breakdown, but lack theoretical richness and demand high computational power compared to the M-estimators. We introduce a novel robust PCA estimator based on the minimum density power divergence estimator. This combines the theoretical strength of the M-estimators and the minimum divergence estimators with a high breakdown guarantee regardless of data dimension. We present a computationally efficient algorithm for this estimate. Our theoretical findings are supported by extensive simulations and comparisons with existing robust PCA methods. We also showcase the proposed algorithm's applicability on two benchmark datasets and a credit card transactions dataset for fraud detection.

This research project delves into the intricacies of road traffic accidents severity in the UK, employing a potent combination of machine learning algorithms, econometric techniques, and traditional statistical methods to analyse longitudinal historical data. Our robust analysis framework includes descriptive, inferential, bivariate, and multivariate methodologies, correlation analysis: Pearson's and Spearman's Rank Correlation Coefficient, multiple and logistic regression models, Multicollinearity Assessment, and Model Validation. In addressing heteroscedasticity or autocorrelation in error terms, we've advanced the precision and reliability of our regression analyses using the Generalized Method of Moments (GMM). Additionally, our application of the Vector Autoregressive (VAR) model and the Autoregressive Integrated Moving Average (ARIMA) models have enabled accurate time-series forecasting. With this approach, we've achieved superior predictive accuracy, marked by a Mean Absolute Scaled Error (MASE) of 0.800 and a Mean Error (ME) of -73.80 compared to a naive forecast.

Large language models (LLMs) have brought significant changes to human society. Softmax regression and residual neural networks (ResNet) are two important techniques in deep learning: they not only serve as significant theoretical components supporting the functionality of LLMs but also are related to many other machine learning and theoretical computer science fields, including but not limited to image classification, object detection, semantic segmentation, and tensors. Previous research works studied these two concepts separately. In this paper, we provide a theoretical analysis of the regression problem: $\| \langle \exp(Ax) + A x , {\bf 1}_n \rangle^{-1} ( \exp(Ax) + Ax ) - b \|_2^2$, where $A$ is a matrix in $\mathbb{R}^{n \times d}$, $b$ is a vector in $\mathbb{R}^n$, and ${\bf 1}_n$ is the $n$-dimensional vector whose entries are all $1$. This regression problem is a unified scheme that combines softmax regression and ResNet, which has never been done before. We derive the gradient, Hessian, and Lipschitz properties of the loss function. The Hessian is shown to be positive semidefinite, and its structure is characterized as the sum of a low-rank matrix and a diagonal matrix. This enables an efficient approximate Newton method. As a result, this unified scheme helps to connect two previously thought unrelated fields and provides novel insight into loss landscape and optimization for emerging over-parameterized neural networks, which is meaningful for future research in deep learning models.

Much of modern data analysis in high dimensions relies on the premise that data, while embedded in a high-dimensional space, lie on or near a submanifold of lower dimension. This allows one to embed the data in a space of lower dimension while preserving much of the essential structure, with benefits including faster computation and data visualization. This lower dimension, hereafter referred to as the intrinsic dimension (ID) of the underlying manifold, often enters as a parameter of the dimension-reduction scheme. For instance, in each of the Johnson-Lindenstrauss-type results for manifolds by [13] and [4] the target dimension depends on the ID. Furthermore, the ID is a parameter of popular dimension reduction methods such as t-SNE [28] and multidimensional scaling [12, 16]. Therefore, it may be beneficial to estimate the ID before running further analysis since compressing the data too much may destroy underlying structure and it may be computationally expensive to re-run algorithms with a new dimension parameter, if such an error is even detectable.

Prediction is a classic challenge in spatial statistics and the inclusion of spatial covariates can greatly improve predictive performance when incorporated into a model with latent spatial effects. It is desirable to develop flexible regression models that allow for nonlinearities and interactions in the covariate structure. Machine learning models have been suggested in the spatial context, allowing for spatial dependence in the residuals, but fail to provide reliable uncertainty estimates. In this paper, we investigate a novel combination of a Gaussian process spatial model and a Bayesian Additive Regression Tree (BART) model. The computational burden of the approach is reduced by combining Markov chain Monte Carlo (MCMC) with the Integrated Nested Laplace Approximation (INLA) technique. We study the performance of the method via simulations and use the model to predict anthropometric responses, collected via household cluster samples in Kenya.

In a recent paper, Ling et al. investigated the over-parametrized Deep Equilibrium Model (DEQ) with ReLU activation. They proved that the gradient descent converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. This paper shows that this fact still holds for DEQs with any general activation that has bounded first and second derivatives. Since the new activation function is generally non-linear, bounding the least eigenvalue of the Gram matrix of the equilibrium point is particularly challenging. To accomplish this task, we need to create a novel population Gram matrix and develop a new form of dual activation with Hermite polynomial expansion.