The success of machine learning models relies heavily on effectively representing high-dimensional data. However, ensuring data representations capture human-understandable concepts remains difficult, often requiring the incorporation of prior knowledge and decomposition of data into multiple subspaces. Traditional linear methods fall short in modeling more than one space, while more expressive deep learning approaches lack interpretability. Here, we introduce Supervised Independent Subspace Principal Component Analysis ($\texttt{sisPCA}$), a PCA extension designed for multi-subspace learning. Leveraging the Hilbert-Schmidt Independence Criterion (HSIC), $\texttt{sisPCA}$ incorporates supervision and simultaneously ensures subspace disentanglement. We demonstrate $\texttt{sisPCA}$'s connections with autoencoders and regularized linear regression and showcase its ability to identify and separate hidden data structures through extensive applications, including breast cancer diagnosis from image features, learning aging-associated DNA methylation changes, and single-cell analysis of malaria infection. Our results reveal distinct functional pathways associated with malaria colonization, underscoring the essentiality of explainable representation in high-dimensional data analysis.

Real-world data is often assumed to lie within a low-dimensional structure embedded in high-dimensional space. In practical settings, we observe only a finite set of samples, forming what we refer to as the sample data subspace. It serves an essential approximation supporting tasks such as dimensionality reduction and generation. A major challenge lies in whether generative models can reliably synthesize samples that stay within this subspace rather than drifting away from the underlying structure. In this work, we provide theoretical insights into this challenge by leveraging Flow Matching models, which transform a simple prior into a complex target distribution via a learned velocity field. By treating the real data distribution as discrete, we derive analytical expressions for the optimal velocity field under a Gaussian prior, showing that generated samples memorize real data points and represent the sample data subspace exactly. To generalize to suboptimal scenarios, we introduce the Orthogonal Subspace Decomposition Network (OSDNet), which systematically decomposes the velocity field into subspace and off-subspace components. Our analysis shows that the off-subspace component decays, while the subspace component generalizes within the sample data subspace, ensuring generated samples preserve both proximity and diversity.

In natural language processing, it is well-established that linear relationships between highdimensional, real-valued vector representations of textual inputs reflect semantic and syntactic patterns. This was motivated in seminal works [4, 5, 6, 7, 8] and extensively validated in word embedding models [9, 10, 11] as well as modern large language models trained for next-token prediction [2, 12, 13, 14, 15, 16, 17, 18, 19]. This ubiquity is puzzling, as different internal representations can produce identical next-token distributions, resulting in distribution-equivalent but internally distinct models. This raises a key question: Are the observed linear properties shared across all models with the same next-token distribution? Our main result is a mathematical proof that, under suitable conditions, certain linear properties hold for either all or none of the equivalent models generating a given next-token distribution. We demonstrate this through three main contributions. The first main contribution (Section 3) is an identifiability result characterizing distributionequivalent next-token predictors. Our result is a generalization of the main theorems by Roeder et al. [3] and Khemakhem et al. [20], relaxing the assumptions of diversity and equal representation dimensionality. This result is of independent interest for research on identifiable representation learning since our analysis is applicable to several discriminative models beyond next-token prediction [3].

Causal discovery with time series data remains a challenging yet increasingly important task across many scientific domains. Convergent cross mapping (CCM) and related methods have been proposed to study time series that are generated by dynamical systems, where traditional approaches like Granger causality are unreliable. However, CCM often yields inaccurate results depending upon the quality of the data. We propose the Tangent Space Causal Inference (TSCI) method for detecting causalities in dynamical systems. TSCI works by considering vector fields as explicit representations of the systems' dynamics and checks for the degree of synchronization between the learned vector fields. The TSCI approach is model-agnostic and can be used as a drop-in replacement for CCM and its generalizations. We first present a basic version of the TSCI algorithm, which is shown to be more effective than the basic CCM algorithm with very little additional computation. We additionally present augmented versions of TSCI that leverage the expressive power of latent variable models and deep learning. We validate our theory on standard systems, and we demonstrate improved causal inference performance across a number of benchmark tasks.

In system analysis and design optimization, multiple computational models are typically available to represent a given physical system. These models can be broadly classified as high-fidelity models, which provide highly accurate predictions but require significant computational resources, and low-fidelity models, which are computationally efficient but less accurate. Multi-fidelity methods integrate high-and low-fidelity models to balance computational cost and predictive accuracy. This perspective paper provides an in-depth overview of the emerging field of machine learning-based multi-fidelity methods, with a particular emphasis on uncertainty quantification and optimization. For uncertainty quantification, a particular focus is on multi-fidelity graph neural networks, compared with multifidelity polynomial chaos expansion. For optimization, our emphasis is on multifidelity Bayesian optimization, offering a unified perspective on multi-fidelity priors and proposing an application strategy when the objective function is an integral or a weighted sum. We highlight the current state of the art, identify critical gaps in the literature, and outline key research opportunities in this evolving field. Keywords multi-fidelity modeling uncertainty quantification Bayesian optimization 1 Introduction When studying a physical system, analysts often have access to multiple computational models. These models are typically classified as either high-fidelity or low-fidelity, depending on their predictive accuracy. High-fidelity models (HFMs) offer precise predictions of the system's behavior, meeting specific accuracy metrics, but they are computationally demanding.

Data increasingly take the form of a multi-way array, or tensor, in several biomedical domains. Such tensors are often incompletely observed. For example, we are motivated by longitudinal microbiome studies in which several timepoints are missing for several subjects. There is a growing literature on missing data imputation for tensors. However, existing methods give a point estimate for missing values without capturing uncertainty. We propose a multiple imputation approach for tensors in a flexible Bayesian framework, that yields realistic simulated values for missing entries and can propagate uncertainty through subsequent analyses. Our model uses efficient and widely applicable conjugate priors for a CANDECOMP/PARAFAC (CP) factorization, with a separable residual covariance structure. This approach is shown to perform well with respect to both imputation accuracy and uncertainty calibration, for scenarios in which either single entries or entire fibers of the tensor are missing. For two microbiome applications, it is shown to accurately capture uncertainty in the full microbiome profile at missing timepoints and used to infer trends in species diversity for the population. Documented R code to perform our multiple imputation approach is available at https://github.com/lockEF/MultiwayImputation .

The problem of regression extrapolation, or out-of-distribution generalization, arises when predictions are required at test points outside the range of the training data. In such cases, the non-parametric guarantees for regression methods from both statistics and machine learning typically fail. Based on the theory of tail dependence, we propose a novel statistical extrapolation principle. After a suitable, data-adaptive marginal transformation, it assumes a simple relationship between predictors and the response at the boundary of the training predictor samples. This assumption holds for a wide range of models, including non-parametric regression functions with additive noise. Our semi-parametric method, progression, leverages this extrapolation principle and offers guarantees on the approximation error beyond the training data range. We demonstrate how this principle can be effectively integrated with existing approaches, such as random forests and additive models, to improve extrapolation performance on out-of-distribution samples.

BART Bayesian Additive Regression Trees (BART) is a nonparametric Bayesian regression method, introduced by Chipman, George, and McCulloch (2006, 2010). It defines a prior distribution over the space of functions by representing them as a sum of binary decision trees, and then specifying a stochastic tree generation process. The posterior is then obtained with Metropolis-Gibbs sampling over the trees. See Hill, Linero, and Murray (2020) for a review, and Daniels, Linero, and Roy (2023, ch. 5) for a textbook treatment. BART's success BART has proven empirically effective, and is gaining popularity (consider, e.g., Tan and Roy 2019). The Atlantic Causal Inference Conference (ACIC) Data Challenge has confirmed BART as one of the best regression methods for causal inference (Dorie et al. 2019; Gruber et al. 2019; Hahn, Dorie, and Murray 2019; Thal and Finucane 2023). Many BART variants have been developed throughout the years, adding features such as variable selection (Linero 2018).

Recent advancements in foundation models have been successfully extended to the time series (TS) domain, facilitated by the emergence of large-scale TS datasets. However, previous efforts have primarily focused on designing model architectures to address explicit heterogeneity among datasets such as various numbers of channels, while often overlooking implicit heterogeneity such as varying dependencies between channels. In this work, we introduce the concept of partial channel dependence (PCD), which enables a more sophisticated adjustment of channel dependencies based on dataset-specific information. To achieve PCD, we propose a channel mask that captures the relationships between channels within a dataset using two key components: 1) a correlation matrix that encodes relative dependencies between channels, and 2) domain parameters that learn the absolute dependencies specific to each dataset, refining the correlation matrix. We validate the effectiveness of PCD across four tasks in TS including forecasting, classification, imputation, and anomaly detection, under diverse settings, including few-shot and zero-shot scenarios with both TS foundation models and single-task models. Code is available at https://github.com/seunghan96/CM.

In graph-based data analysis, $k$-nearest neighbor ($k$NN) graphs are widely used due to their adaptivity to local data densities. Allowing weighted edges in the graph, the kernelized graph affinity provides a more general type of $k$NN graph where the $k$NN distance is used to set the kernel bandwidth adaptively. In this work, we consider a general class of $k$NN graph where the graph affinity is $W_{ij} = \epsilon^{-d/2} \; k_0 ( \| x_i - x_j \|^2 / \epsilon \phi( \widehat{\rho}(x_i), \widehat{\rho}(x_j) )^2 ) $, with $\widehat{\rho}(x)$ being the (rescaled) $k$NN distance at the point $x$, $\phi$ a symmetric bi-variate function, and $k_0$ a non-negative function on $[0,\infty)$. Under the manifold data setting, where $N$ i.i.d. samples $x_i$ are drawn from a density $p$ on a $d$-dimensional unknown manifold embedded in a high dimensional Euclidean space, we prove the point-wise convergence of the $k$NN graph Laplacian to the limiting manifold operator (depending on $p$) at the rate of $O(N^{-2/(d+6)}\,)$, up to a log factor, when $k_0$ and $\phi$ have $C^3$ regularity and satisfy other technical conditions. This fast rate is obtained when $\epsilon \sim N^{-2/(d+6)}\,$ and $k \sim N^{6/(d+6)}\,$, both at the optimal order to balance the theoretical bias and variance errors. When $k_0$ and $\phi$ have lower regularities, including when $k_0$ is a compactly supported function as in the standard $k$NN graph, the convergence rate degenerates to $O(N^{-1/(d+4)}\,)$. Our improved convergence rate is based on a refined analysis of the $k$NN estimator, which can be of independent interest. We validate our theory by numerical experiments on simulated data.