Fr\'echet regression extends classical regression methods to non-Euclidean metric spaces, enabling the analysis of data relationships on complex structures such as manifolds and graphs. This work establishes a rigorous theoretical analysis for Fr\'echet regression through the lens of comparison geometry which leads to important considerations for its use in practice. The analysis provides key results on the existence, uniqueness, and stability of the Fr\'echet mean, along with statistical guarantees for nonparametric regression, including exponential concentration bounds and convergence rates. Additionally, insights into angle stability reveal the interplay between curvature of the manifold and the behavior of the regression estimator in these non-Euclidean contexts. Empirical experiments validate the theoretical findings, demonstrating the effectiveness of proposed hyperbolic mappings, particularly for data with heteroscedasticity, and highlighting the practical usefulness of these results.

Grokking, or delayed generalization, is an intriguing learning phenomenon where test set loss decreases sharply only after a model's training set loss has converged. This challenges conventional understanding of the training dynamics in deep learning networks. In this paper, we formalize and investigate grokking, highlighting that a key factor in its emergence is a distribution shift between training and test data. We introduce two synthetic datasets specifically designed to analyze grokking. One dataset examines the impact of limited sampling, and the other investigates transfer learning's role in grokking. By inducing distribution shifts through controlled imbalanced sampling of sub-categories, we systematically reproduce the phenomenon, demonstrating that while small-sampling is strongly associated with grokking, it is not its cause. Instead, small-sampling serves as a convenient mechanism for achieving the necessary distribution shift. We also show that when classes form an equivariant map, grokking can be explained by the model's ability to learn from similar classes or sub-categories. Unlike earlier work suggesting that grokking primarily arises from high regularization and sparse data, we demonstrate that it can also occur with dense data and minimal hyper-parameter tuning. Our findings deepen the understanding of grokking and pave the way for developing better stopping criteria in future training processes.

Printed Electronics (PE) technology has emerged as a promising alternative to silicon-based computing. It offers attractive properties such as on-demand ultra-low-cost fabrication, mechanical flexibility, and conformality. However, PE are governed by large feature sizes, prohibiting the realization of complex printed Machine Learning (ML) classifiers. Leveraging PE's ultra-low non-recurring engineering and fabrication costs, designers can fully customize hardware to a specific ML model and dataset, significantly reducing circuit complexity. Despite significant advancements, state-of-the-art solutions achieve area efficiency at the expense of considerable accuracy loss. Our work mitigates this by designing area- and power-efficient printed ML classifiers with little to no accuracy degradation. Specifically, we introduce the first sequential Support Vector Machine (SVM) classifiers, exploiting the hardware efficiency of bespoke control and storage units and a single Multiply-Accumulate compute engine. Our SVMs yield on average 6x lower area and 4.6% higher accuracy compared to the printed state of the art.

Multiview feature learning aims to learn discriminative features by integrating the distinct information in each view. However, most existing methods still face significant challenges in learning view-consistency features, which are crucial for effective multiview learning. Motivated by the theories of CCA and contrastive learning in multiview feature learning, we propose the hierarchical consensus network (HCN) in this paper. The HCN derives three consensus indices for capturing the hierarchical consensus across views, which are classifying consensus, coding consensus, and global consensus, respectively. Specifically, classifying consensus reinforces class-level correspondence between views from a CCA perspective, while coding consensus closely resembles contrastive learning and reflects contrastive comparison of individual instances. Global consensus aims to extract consensus information from two perspectives simultaneously. By enforcing the hierarchical consensus, the information within each view is better integrated to obtain more comprehensive and discriminative features. The extensive experimental results obtained on four multiview datasets demonstrate that the proposed method significantly outperforms several state-of-the-art methods.

Symmetry arises often when learning from high dimensional data. For example, data sets consisting of point clouds, graphs, and unordered sets appear routinely in contemporary applications, and exhibit rich underlying symmetries. Understanding the benefits of symmetry on the statistical and numerical efficiency of learning algorithms is an active area of research. In this work, we show that symmetry has a pronounced impact on the rank of kernel matrices. Specifically, we compute the rank of a polynomial kernel of fixed degree that is invariant under various groups acting independently on its two arguments. In concrete circumstances, including the three aforementioned examples, symmetry dramatically decreases the rank making it independent of the data dimension. In such settings, we show that a simple regression procedure is minimax optimal for estimating an invariant polynomial from finitely many samples drawn across different dimensions. We complete the paper with numerical experiments that illustrate our findings.

Multi-index models provide a popular framework to investigate the learnability of functions with low-dimensional structure and, also due to their connections with neural networks, they have been object of recent intensive study. In this paper, we focus on recovering the subspace spanned by the signals via spectral estimators -- a family of methods that are routinely used in practice, often as a warm-start for iterative algorithms. Our main technical contribution is a precise asymptotic characterization of the performance of spectral methods, when sample size and input dimension grow proportionally and the dimension $p$ of the space to recover is fixed. Specifically, we locate the top-$p$ eigenvalues of the spectral matrix and establish the overlaps between the corresponding eigenvectors (which give the spectral estimators) and a basis of the signal subspace. Our analysis unveils a phase transition phenomenon in which, as the sample complexity grows, eigenvalues escape from the bulk of the spectrum and, when that happens, eigenvectors recover directions of the desired subspace. The precise characterization we put forward enables the optimization of the data preprocessing, thus allowing to identify the spectral estimator that requires the minimal sample size for weak recovery.

The machine learning (ML) training over disparate data sources traditionally involves materialization, which can impose substantial time and space overhead due to data movement and replication. Factorized learning, which leverages direct computation on disparate sources through linear algebra (LA) rewriting, has emerged as a viable alternative to improve computational efficiency. However, the adaptation of factorized learning to leverage the full capabilities of modern LA-friendly hardware like GPUs has been limited, often requiring manual intervention for algorithm compatibility. This paper introduces Ilargi, a novel factorized learning framework that utilizes matrix-represented data integration (DI) metadata to facilitate automatic factorization across CPU and GPU environments without the need for costly relational joins. Ilargi incorporates an ML-based cost estimator to intelligently selects between factorization and materialization based on data properties, algorithm complexity, hardware environments, and their interactions. This strategy ensures up to 8.9x speedups on GPUs and achieves over 20% acceleration in batch ML training workloads, thereby enhancing the practicability of ML training across diverse data integration scenarios and hardware platforms. To our knowledge, this work is the very first effort in GPU-compatible factorized learning.

Gradient Boosting Decision Tree (GBDT) is one of the most popular machine learning models in various applications. However, in the traditional settings, all data should be simultaneously accessed in the training procedure: it does not allow to add or delete any data instances after training. In this paper, we propose an efficient online learning framework for GBDT supporting both incremental and decremental learning. To the best of our knowledge, this is the first work that considers an in-place unified incremental and decremental learning on GBDT. To reduce the learning cost, we present a collection of optimizations for our framework, so that it can add or delete a small fraction of data on the fly. We theoretically show the relationship between the hyper-parameters of the proposed optimizations, which enables trading off accuracy and cost on incremental and decremental learning. The backdoor attack results show that our framework can successfully inject and remove backdoor in a well-trained model using incremental and decremental learning, and the empirical results on public datasets confirm the effectiveness and efficiency of our proposed online learning framework and optimizations.

Layer-wise preconditioning methods are a family of memory-efficient optimization algorithms that introduce preconditioners per axis of each layer's weight tensors. These methods have seen a recent resurgence, demonstrating impressive performance relative to entry-wise ("diagonal") preconditioning methods such as Adam(W) on a wide range of neural network optimization tasks. Complementary to their practical performance, we demonstrate that layer-wise preconditioning methods are provably necessary from a statistical perspective. To showcase this, we consider two prototypical models, linear representation learning and single-index learning, which are widely used to study how typical algorithms efficiently learn useful features to enable generalization. In these problems, we show SGD is a suboptimal feature learner when extending beyond ideal isotropic inputs $\mathbf{x} \sim \mathsf{N}(\mathbf{0}, \mathbf{I})$ and well-conditioned settings typically assumed in prior work. We demonstrate theoretically and numerically that this suboptimality is fundamental, and that layer-wise preconditioning emerges naturally as the solution. We further show that standard tools like Adam preconditioning and batch-norm only mildly mitigate these issues, supporting the unique benefits of layer-wise preconditioning.

Performing gradient descent in a wide neural network is equivalent to computing the posterior mean of a Gaussian Process with the Neural Tangent Kernel (NTK-GP), for a specific choice of prior mean and with zero observation noise. However, existing formulations of this result have two limitations: i) the resultant NTK-GP assumes no noise in the observed target variables, which can result in suboptimal predictions with noisy data; ii) it is unclear how to extend the equivalence to an arbitrary prior mean, a crucial aspect of formulating a well-specified model. To address the first limitation, we introduce a regularizer into the neural network's training objective, formally showing its correspondence to incorporating observation noise into the NTK-GP model. To address the second, we introduce a \textit{shifted network} that enables arbitrary prior mean functions. This approach allows us to perform gradient descent on a single neural network, without expensive ensembling or kernel matrix inversion. Our theoretical insights are validated empirically, with experiments exploring different values of observation noise and network architectures.