This paper introduces torchsom, an open-source Python library that provides a reference implementation of the Self-Organizing Map (SOM) in PyTorch. This package offers three main features: (i) dimensionality reduction, (ii) clustering, and (iii) friendly data visualization. It relies on a PyTorch backend, enabling (i) fast and efficient training of SOMs through GPU acceleration, and (ii) easy and scalable integrations with PyTorch ecosystem. Moreover, torchsom follows the scikit-learn API for ease of use and extensibility.

Forecasting has undergone a profound transformation in the 21st century, driven by advancements in methodology, computational power, and data availability. The origins of this transformation can be traced back to the late 1990s, when the field of economics began grappling with the challenges and opportunities presented by what is now termed "Big Data"--the availability of large datasets with numerous predictors. This period marked the emergence of systematic efforts to develop tools capable of addressing the high-dimensional nature of these datasets with seminal contributions by Frank Diebold, Mario Forni, Marc Hallin, Marco Lippi, Lucrezia Reichlin, Jim Stock, and Mark Watson (see Reichlin, 2003; Watson, 2003; Diebold, 2003). They laid the foundation for a new era of forecasting, as presented at the World Meeting of the Econometric Society in the summer of 2000 (for a survey see De Mol et al., 2017; Diebold, 2021). Since these early contributions, the field of macroeconomic forecasting has experienced significant progress.

This paper develops a finite-sample statistical theory for in-context learning (ICL), analyzed within a meta-learning framework that accommodates mixtures of diverse task types. We introduce a principled risk decomposition that separates the total ICL risk into two orthogonal components: Bayes Gap and Posterior Variance. The Bayes Gap quantifies how well the trained model approximates the Bayes-optimal in-context predictor. For a uniform-attention Transformer, we derive a non-asymptotic upper bound on this gap, which explicitly clarifies the dependence on the number of pretraining prompts and their context length. The Posterior Variance is a model-independent risk representing the intrinsic task uncertainty. Our key finding is that this term is determined solely by the difficulty of the true underlying task, while the uncertainty arising from the task mixture vanishes exponentially fast with only a few in-context examples. Together, these results provide a unified view of ICL: the Transformer selects the optimal meta-algorithm during pretraining and rapidly converges to the optimal algorithm for the true task at test time.

We present a theoretical framework that extends classical information theory to finite and structured systems by redefining redundancy as a fundamental property of information organization rather than inefficiency. In this framework, redundancy is expressed as a general family of informational divergences that unifies multiple classical measures, such as mutual information, chi-squared dependence, and spectral redundancy, under a single geometric principle. This reveals that these traditional quantities are not isolated heuristics but projections of a shared redundancy geometry. The theory further predicts that redundancy is bounded both above and below, giving rise to an optimal equilibrium that balances over-compression (loss of structure) and over-coupling (collapse). While classical communication theory favors minimal redundancy for transmission efficiency, finite and structured systems, such as those underlying real-world learning, achieve maximal stability and generalization near this equilibrium. Experiments with masked autoencoders are used to illustrate and verify this principle: the model exhibits a stable redundancy level where generalization peaks. Together, these results establish redundancy as a measurable and tunable quantity that bridges the asymptotic world of communication and the finite world of learning.

Random forest is an important method for ML applications due to its broad outperformance over competing methods for structured tabular data. We propose a method for transfer learning in nonparametric regression using a centered random forest (CRF) with distance covariance-based feature weights, assuming the unknown source and target regression functions are different for a few features (sparsely different). Our method first obtains residuals from predicting the response in the target domain using a source domain-trained CRF. Then, we fit another CRF to the residuals, but with feature splitting probabilities proportional to the sample distance covariance between the features and the residuals in an independent sample. We derive an upper bound on the mean square error rate of the procedure as a function of sample sizes and difference dimension, theoretically demonstrating transfer learning benefits in random forests. In simulations, we show that the results obtained for the CRFs also hold numerically for the standard random forest (SRF) method with data-driven feature split selection. Beyond transfer learning, our results also show the benefit of distance-covariance-based weights on the performance of RF in some situations. Our method shows significant gains in predicting the mortality of ICU patients in smaller-bed target hospitals using a large multi-hospital dataset of electronic health records for 200,000 ICU patients.

We provide a unified algorithmic framework for ensemble sampling in nonlinear contextual bandits and develop corresponding regret bounds for two most common nonlinear contextual bandit settings: Generalized Linear Ensemble Sampling (\texttt{GLM-ES}) for generalized linear bandits and Neural Ensemble Sampling (\texttt{Neural-ES}) for neural contextual bandits. Both methods maintain multiple estimators for the reward model parameters via maximum likelihood estimation on randomly perturbed data. We prove high-probability frequentist regret bounds of $\mathcal{O}(d^{3/2} \sqrt{T} + d^{9/2})$ for \texttt{GLM-ES} and $\mathcal{O}(\widetilde{d} \sqrt{T})$ for \texttt{Neural-ES}, where $d$ is the dimension of feature vectors, $\widetilde{d}$ is the effective dimension of a neural tangent kernel matrix, and $T$ is the number of rounds. These regret bounds match the state-of-the-art results of randomized exploration algorithms in nonlinear contextual bandit settings. In the theoretical analysis, we introduce techniques that address challenges specific to nonlinear models. Practically, we remove fixed-time horizon assumptions by developing anytime versions of our algorithms, suitable when $T$ is unknown. Finally, we empirically evaluate \texttt{GLM-ES}, \texttt{Neural-ES}, and their anytime variants, demonstrating strong performance. Overall, our results establish ensemble sampling as a provable and practical randomized exploration approach for nonlinear contextual bandits.

Quantized neural network training optimizes a discrete, non-differentiable objective. The straight-through estimator (STE) enables backpropagation through surrogate gradients and is widely used. While previous studies have primarily focused on the properties of surrogate gradients and their convergence, the influence of quantization hyperparameters, such as bit width and quantization range, on learning dynamics remains largely unexplored. We theoretically show that in the high-dimensional limit, STE dynamics converge to a deterministic ordinary differential equation. This reveals that STE training exhibits a plateau followed by a sharp drop in generalization error, with plateau length depending on the quantization range. A fixed-point analysis quantifies the asymptotic deviation from the unquantized linear model. We also extend analytical techniques for stochastic gradient descent to nonlinear transformations of weights and inputs.

We study the task of noiseless linear regression under Gaussian covariates in the presence of additive oblivious contamination. Specifically, we are given i.i.d.\ samples from a distribution $(x, y)$ on $\mathbb{R}^d \times \mathbb{R}$ with $x \sim \mathcal{N}(0,\mathbf{I}_d)$ and $y = x^\top β+ z$, where $z$ is drawn independently of $x$ from an unknown distribution $E$. Moreover, $z$ satisfies $\mathbb{P}_E[z = 0] = α>0$. The goal is to accurately recover the regressor $β$ to small $\ell_2$-error. Ignoring computational considerations, this problem is known to be solvable using $O(d/α)$ samples. On the other hand, the best known polynomial-time algorithms require $Ω(d/α^2)$ samples. Here we provide formal evidence that the quadratic dependence in $1/α$ is inherent for efficient algorithms. Specifically, we show that any efficient Statistical Query algorithm for this task requires VSTAT complexity at least $\tildeΩ(d^{1/2}/α^2)$.

Deep learning based weather forecasting (DLWF) models leverage past weather observations to generate future forecasts, supporting a wide range of downstream tasks, including tropical cyclone (TC) trajectory prediction. In this paper, we investigate their vulnerability to adversarial attacks, where subtle perturbations to the upstream weather forecasts can alter the downstream TC trajectory predictions. Although research on adversarial attacks in DLWF models has grown recently, generating perturbed upstream forecasts that reliably steer downstream output toward attacker-specified trajectories remains a challenge. First, conventional TC detection systems are opaque, non-differentiable black boxes, making standard gradient-based attacks infeasible. Second, the extreme rarity of TC events leads to severe class imbalance problem, making it difficult to develop efficient attack methods that will produce the attacker's target trajectories. Furthermore, maintaining physical consistency in adversarially generated forecasts presents another significant challenge. To overcome these limitations, we propose Cyc-Attack, a novel method that perturbs the upstream forecasts of DLWF models to generate adversarial trajectories. First, we pre-train a differentiable surrogate model to approximate the TC detector's output, enabling the construction of gradient-based attacks. Cyc-Attack also employs skewness-aware loss function with kernel dilation strategy to address the imbalance problem. Finally, a distance-based gradient weighting scheme and regularization are used to constrain the perturbations and eliminate spurious trajectories to ensure the adversarial forecasts are realistic and not easily detectable.

Incomplete multi-view spectral clustering generalizes spectral clustering to multi-view data and simultaneously realizes the partition of multi-view data with missing views. For this category of method, K-means algorithm needs to be performed to generate the clustering result after the procedure of feature extraction. More importantly, the connectivity of samples reflected by the clustering result is not utilized effectively. To overcome these defects, we propose Clustering Result re-Guided Incomplete Multi-view Spectral Clustering (CRG_IMSC). CRG_IMSC obtains the clustering result directly by imposing nonnegative constraint to the extracted feature. Furthermore, it constructs the connectivity matrix according to the result of spectral clustering, and minimizes the residual of self-representation based on the connectivity matrix. A novel iterative algorithm using multiplicative update is developed to solve the optimization problem of CRG_IMSC, and its convergence is proved rigorously. On benchmark datasets, for multi-view data, CRG_IMSC performs better than state-of-the-art clustering methods, and the experimental results also demonstrate the convergence of CRG_IMSC algorithm.