Tensor decomposition (TD) is essential for analyzing high-dimensional sparse data, yet its irregular computations and memory-access patterns pose major performance challenges on modern parallel processors. Prior works rely on expert-designed sparse tensor formats that fail to adapt to irregular tensor shapes and/or highly variable data distributions. We present the reinforcement-learned adaptive tensor encoding (ReLATE) framework, a novel learning-augmented method that automatically constructs efficient sparse tensor representations without labeled training samples. ReLATE employs an autonomous agent that discovers optimized tensor encodings through direct interaction with the TD environment, leveraging a hybrid model-free and model-based algorithm to learn from both real and imagined actions. Moreover, ReLATE introduces rule-driven action masking and dynamics-informed action filtering mechanisms that ensure functionally correct tensor encoding with bounded execution time, even during early learning stages. By automatically adapting to both irregular tensor shapes and data distributions, ReLATE generates sparse tensor representations that consistently outperform expert-designed formats across diverse sparse tensor data sets, achieving up to 2X speedup compared to the best sparse format, with a geometric-mean speedup of 1.4-1.46X.

The exponential growth of artificial intelligence (AI) and machine learning (ML) applications has necessitated the development of efficient storage solutions for vector and tensor data. This paper presents a novel approach for tensor storage in a Lakehouse architecture using Delta Lake. By adopting the multidimensional array storage strategy from array databases and sparse encoding methods to Delta Lake tables, experiments show that this approach has demonstrated notable improvements in both space and time efficiencies when compared to traditional serialization of tensors. These results provide valuable insights for the development and implementation of optimized vector and tensor storage solutions in data-intensive applications, contributing to the evolution of efficient data management practices in AI and ML domains in cloud-native environments

We study the approximation of functions by tensor networks (TNs). We show that Lebesgue $L^p$-spaces in one dimension can be identified with tensor product spaces of arbitrary order through tensorization. We use this tensor product structure to define subsets of $L^p$ of rank-structured functions of finite representation complexity. These subsets are then used to define different approximation classes of tensor networks, associated with different measures of complexity. These approximation classes are shown to be quasi-normed linear spaces. We study some elementary properties and relationships of said spaces. In part II of this work, we will show that classical smoothness (Besov) spaces are continuously embedded into these approximation classes. We will also show that functions in these approximation classes do not possess any Besov smoothness, unless one restricts the depth of the tensor networks. The results of this work are both an analysis of the approximation spaces of TNs and a study of the expressivity of a particular type of neural networks (NN) -- namely feed-forward sum-product networks with sparse architecture. The input variables of this network result from the tensorization step, interpreted as a particular featuring step which can also be implemented with a neural network with a specific architecture. We point out interesting parallels to recent results on the expressivity of rectified linear unit (ReLU) networks -- currently one of the most popular type of NNs.

Tensorial Convolutional Neural Networks (TCNNs) have attracted much research attention for their power in reducing model parameters or enhancing the generalization ability. However, exploration of TCNNs is hindered even from weight initialization methods. To be specific, general initialization methods, such as Xavier or Kaiming initialization, usually fail to generate appropriate weights for TCNNs. Meanwhile, although there are ad-hoc approaches for specific architectures (e.g., Tensor Ring Nets), they are not applicable to TCNNs with other tensor decomposition methods (e.g., CP or Tucker decomposition). To address this problem, we propose a universal weight initialization paradigm, which generalizes Xavier and Kaiming methods and can be widely applicable to arbitrary TCNNs. Specifically, we first present the Reproducing Transformation to convert the backward process in TCNNs to an equivalent convolution process. Then, based on the convolution operators in the forward and backward processes, we build a unified paradigm to control the variance of features and gradients in TCNNs. Thus, we can derive fan-in and fan-out initialization for various TCNNs. We demonstrate that our paradigm can stabilize the training of TCNNs, leading to faster convergence and better results.

Low-rank tensor decomposition is one of the most effective approaches to reduce the memory and computing requirements of large-size neural networks, enabling their efficient deployment on various hardware platforms. While post-training tensor compression can greatly reduce the cost of inference, uncompressed training still consumes excessive hardware resources, run-time and energy. It is highly desirable to directly train a compact low-rank tensorized model from scratch with a low memory and computational cost. However, this is a very challenging task because it is hard to determine a proper tensor rank a priori, which controls the model complexity and compression ratio in the training process. This paper presents a novel end-to-end framework for low-rank tensorized training of neural networks. We first develop a flexible Bayesian model that can handle various low-rank tensor formats (e.g., CP, Tucker, tensor train and tensor-train matrix) that compress neural network parameters in training. This model can automatically determine the tensor ranks inside a nonlinear forward model, which is beyond the capability of existing Bayesian tensor methods. We further develop a scalable stochastic variational inference solver to estimate the posterior density of large-scale problems in training. Our work provides the first general-purpose rank-adaptive framework for end-to-end tensorized training. Our numerical results on various neural network architectures show orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the training process.

In this paper, we propose and analyze a model selection method for tree tensor networks in an empirical risk minimization framework. Tree tensor networks, or tree-based tensor formats, are prominent model classes for the approximation of high-dimensional functions in numerical analysis and data science. They correspond to sum-product neural networks with a sparse connectivity associated with a dimension partition tree $T$, widths given by a tuple $r$ of tensor ranks, and multilinear activation functions (or units). The approximation power of these model classes has been proved to be near-optimal for classical smoothness classes. However, in an empirical risk minimization framework with a limited number of observations, the dimension tree $T$ and ranks $r$ should be selected carefully to balance estimation and approximation errors. In this paper, we propose a complexity-based model selection strategy \`a la Barron, Birg\'e, Massart. Given a family of model classes, with different trees, ranks and tensor product feature spaces, a model is selected by minimizing a penalized empirical risk, with a penalty depending on the complexity of the model class. After deriving bounds of the metric entropy of tree tensor networks with bounded parameters, we deduce a form of the penalty from bounds on suprema of empirical processes. This choice of penalty yields a risk bound for the predictor associated with the selected model. For classical smoothness spaces, we show that the proposed strategy is minimax optimal in a least-squares setting. In practice, the amplitude of the penalty is calibrated with a slope heuristics method. Numerical experiments in a least-squares regression setting illustrate the performance of the strategy for the approximation of multivariate functions and univariate functions identified with tensors by tensorization (quantization).

We consider the problem of the estimation of a high-dimensional probability distribution using model classes of functions in tree-based tensor formats, a particular case of tensor networks associated with a dimension partition tree. The distribution is assumed to admit a density with respect to a product measure, possibly discrete for handling the case of discrete random variables. After discussing the representation of classical model classes in tree-based tensor formats, we present learning algorithms based on empirical risk minimization using a $L^2$ contrast. These algorithms exploit the multilinear parametrization of the formats to recast the nonlinear minimization problem into a sequence of empirical risk minimization problems with linear models. A suitable parametrization of the tensor in tree-based tensor format allows to obtain a linear model with orthogonal bases, so that each problem admits an explicit expression of the solution and cross-validation risk estimates. These estimations of the risk enable the model selection, for instance when exploiting sparsity in the coefficients of the representation. A strategy for the adaptation of the tensor format (dimension tree and tree-based ranks) is provided, which allows to discover and exploit some specific structures of high-dimensional probability distributions such as independence or conditional independence. We illustrate the performances of the proposed algorithms for the approximation of classical probabilistic models (such as Gaussian distribution, graphical models, Markov chain).

This paper is concerned with the approximation of high-dimensional functions in a statistical learning setting, by empirical risk minimization over model classes of functions in tree-based tensor format. These are particular classes of rank-structured functions that can be seen as deep neural networks with a sparse architecture related to the tree and multilinear activation functions. For learning in a given model class, we exploit the fact that tree-based tensor formats are multilinear models and recast the problem of risk minimization over a nonlinear set into a succession of learning problems with linear models. Suitable changes of representation yield numerically stable learning problems and allow to exploit sparsity. For high-dimensional problems or when only a small data set is available, the selection of a good model class is a critical issue. For a given tree, the selection of the tuple of tree-based ranks that minimize the risk is a combinatorial problem. Here, we propose a rank adaptation strategy which provides in practice a good convergence of the risk as a function of the model class complexity. Finding a good tree is also a combinatorial problem, which can be related to the choice of a particular sparse architecture for deep neural networks. Here, we propose a stochastic algorithm for minimizing the complexity of the representation of a given function over a class of trees with a given arity, allowing changes in the topology of the tree. This tree optimization algorithm is then included in a learning scheme that successively adapts the tree and the corresponding tree-based ranks. Contrary to classical learning algorithms for nonlinear model classes, the proposed algorithms are numerically stable, reliable, and require only a low level expertise of the user.