We present a unified computational theory of perception and memory. In our model, perception, episodic memory, and semantic memory are realized by different functional and operational modes of the oscillating interactions between an index layer and a representation layer in a bilayer tensor network (BTN). The memoryless semantic {representation layer} broadcasts information. In cognitive neuroscience, it would be the "mental canvas", or the "global workspace" and reflects the cognitive brain state. The symbolic {index layer} represents concepts and past episodes, whose semantic embeddings are implemented in the connection weights between both layers. In addition, we propose a {working memory layer} as a processing center and information buffer. Episodic and semantic memory realize memory-based reasoning, i.e., the recall of relevant past information to enrich perception, and are personalized to an agent's current state, as well as to an agent's unique memories. Episodic memory stores and retrieves past observations and provides provenance and context. Recent episodic memory enriches perception by the retrieval of perceptual experiences, which provide the agent with a sense about the here and now: to understand its own state, and the world's semantic state in general, the agent needs to know what happened recently, in recent scenes, and on recently perceived entities. Remote episodic memory retrieves relevant past experiences, contributes to our conscious self, and, together with semantic memory, to a large degree defines who we are as individuals.

The solution of multistage stochastic linear problems (MSLP) represents a challenge for many applications. Long-term hydrothermal dispatch planning (LHDP) materializes this challenge in a real-world problem that affects electricity markets, economies, and natural resources worldwide. No closed-form solutions are available for MSLP and the definition of non-anticipative policies with high-quality out-of-sample performance is crucial. Linear decision rules (LDR) provide an interesting simulation-based framework for finding high-quality policies to MSLP through two-stage stochastic models. In practical applications, however, the number of parameters to be estimated when using an LDR may be close or higher than the number of scenarios, thereby generating an in-sample overfit and poor performances in out-of-sample simulations. In this paper, we propose a novel regularization scheme for LDR based on the AdaLASSO (adaptive least absolute shrinkage and selection operator). The goal is to use the parsimony principle as largely studied in high-dimensional linear regression models to obtain better out-of-sample performance for an LDR applied to MSLP. Computational experiments show that the overfit threat is non-negligible when using the classical non-regularized LDR to solve MSLP. For the LHDP problem, our analysis highlights the following benefits of the proposed framework in comparison to the non-regularized benchmark: 1) significant reductions in the number of non-zero coefficients (model parsimony), 2) substantial cost reductions in out-of-sample evaluations, and 3) improved spot-price profiles.

Here, we show that the robust overfitting shall be viewed as the early part of an epoch-wise double descent -- the robust test error will start to decrease again after training the model for a considerable number of epochs. Inspired by our observations, we further advance the analyses of double descent to understand robust overfitting better. In standard training, double descent has been shown to be a result of label flipping noise. However, this reasoning is not applicable in our setting, since adversarial perturbations are believed not to change the label. Going beyond label flipping noise, we propose to measure the mismatch between the assigned and (unknown) true label distributions, denoted as \emph{implicit label noise}. We show that the traditional labeling of adversarial examples inherited from their clean counterparts will lead to implicit label noise. Towards better labeling, we show that predicted distribution from a classifier, after scaling and interpolation, can provably reduce the implicit label noise under mild assumptions. In light of our analyses, we tailored the training objective accordingly to effectively mitigate the double descent and verified its effectiveness on three benchmark datasets.

Kernel two-sample tests have been widely used and the development of efficient methods for high-dimensional large-scale data is gaining more and more attention as we are entering the big data era. However, existing methods, such as the maximum mean discrepancy (MMD) and recently proposed kernel-based tests for large-scale data, are computationally intensive to implement and/or ineffective for some common alternatives for high-dimensional data. In this paper, we propose a new test that exhibits high power for a wide range of alternatives. Moreover, the new test is more robust to high dimensions than existing methods and does not require optimization procedures for the choice of kernel bandwidth and other parameters by data splitting. Numerical studies show that the new approach performs well in both synthetic and real world data.

Machine Learning models are deployed across a wide range of industries, performing a wide range of tasks. Tracking these models and ensuring they behave appropriately is becoming increasingly difficult as the number of deployed models increases. There are also new regulatory burdens for ML systems which affect human lives, requiring a link between a model and its training data in high-risk situations. Current ML monitoring systems often provide provenance and experiment tracking as a layer on top of an ML library, allowing room for imperfect tracking and skew between the tracked object and the metadata. In this paper we introduce Tribuo, a Java ML library that integrates model training, inference, strong type-safety, runtime checking, and automatic provenance recording into a single framework. All Tribuo's models and evaluations record the full processing pipeline for input data, along with the training algorithms, hyperparameters and data transformation steps automatically. The provenance lives inside the model object and can be persisted separately using common markup formats. Tribuo implements many popular ML algorithms for classification, regression, clustering, multi-label classification and anomaly detection, along with interfaces to XGBoost, TensorFlow and ONNX Runtime. Tribuo's source code is available at https://github.com/oracle/tribuo under an Apache 2.0 license with documentation and tutorials available at https://tribuo.org.

Often in statistics and machine learning we are required to partition a dataset, e.g., when (i) splitting a dataset for training and testing, (ii) subsampling from Big Data for conducting tractable statistical analysis or to save storage space, (iii) generating multiple splits of a dataset for divide-and-conquer procedures to act upon, and (iv) creating k-fold cross validation sets for model tuning and validation. For this purpose, we propose a novel method named Twinning that can be used for partitioning a dataset into statistically similar sets. Twinning is motivated from the recent work on optimal data splitting for model validation, by Joseph and Vakayil (2021). For model validation, the common practice is to randomly split the dataset into training and testing sets, e.g., for an 80-20 split, 20% of the dataset is selected randomly for testing, while the remaining 80% is used for training the model. It is easy to see that such random splitting can plausibly give rise to pathological splits, wherein the training and testing sets cover roughly disjoint regions of the feature space, thereby resulting in poor testing performance of the model.

Finding the optimal configuration of parameters in ResNet is a nonconvex minimization problem, but first-order methods nevertheless find the global optimum in the overparameterized regime. We study this phenomenon with mean-field analysis, by translating the training process of ResNet to a gradient-flow partial differential equation (PDE) and examining the convergence properties of this limiting process. The activation function is assumed to be $2$-homogeneous or partially $1$-homogeneous; the regularized ReLU satisfies the latter condition. We show that if the ResNet is sufficiently large, with depth and width depending algebraically on the accuracy and confidence levels, first-order optimization methods can find global minimizers that fit the training data.

Including covariant information, such as position, force, velocity or spin is important in many tasks in computational physics and chemistry. We introduce Steerable E(3) Equivariant Graph Neural Networks (SEGNNs) that generalise equivariant graph networks, such that node and edge attributes are not restricted to invariant scalars, but can contain covariant information, such as vectors or tensors. This model, composed of steerable MLPs, is able to incorporate geometric and physical information in both the message and update functions. Through the definition of steerable node attributes, the MLPs provide a new class of activation functions for general use with steerable feature fields. We discuss ours and related work through the lens of equivariant non-linear convolutions, which further allows us to pin-point the successful components of SEGNNs: non-linear message aggregation improves upon classic linear (steerable) point convolutions; steerable messages improve upon recent equivariant graph networks that send invariant messages. We demonstrate the effectiveness of our method on several tasks in computational physics and chemistry and provide extensive ablation studies.

We propose a relative entropy gradient sampler (REGS) for sampling from unnormalized distributions. REGS is a particle method that seeks a sequence of simple nonlinear transforms iteratively pushing the initial samples from a reference distribution into the samples from an unnormalized target distribution. To determine the nonlinear transforms at each iteration, we consider the Wasserstein gradient flow of relative entropy. This gradient flow determines a path of probability distributions that interpolates the reference distribution and the target distribution. It is characterized by an ODE system with velocity fields depending on the density ratios of the density of evolving particles and the unnormalized target density. To sample with REGS, we need to estimate the density ratios and simulate the ODE system with particle evolution. We propose a novel nonparametric approach to estimating the logarithmic density ratio using neural networks. Extensive simulation studies on challenging multimodal 1D and 2D mixture distributions and Bayesian logistic regression on real datasets demonstrate that the REGS outperforms the state-of-the-art sampling methods included in the comparison.

Network-valued data are encountered in a wide range of applications, and pose challenges in learning due to their complex structure and absence of vertex correspondence. Typical examples of such problems include classification or grouping of protein structures and social networks. Various methods, ranging from graph kernels to graph neural networks, have been proposed that achieve some success in graph classification problems. However, most methods have limited theoretical justification, and their applicability beyond classification remains unexplored. In this work, we propose methods for clustering multiple graphs, without vertex correspondence, that are inspired by the recent literature on estimating graphons-- symmetric functions corresponding to infinite vertex limit of graphs. We propose a novel graph distance based on sorting-and-smoothing graphon estimators. Using the proposed graph distance, we present two clustering algorithms and show that they achieve state-of-the-art results. We prove the statistical consistency of both algorithms under Lipschitz assumptions on the graph degrees. We further study the applicability of the proposed distance for graph two-sample testing problems. Machine learning on graphs has evolved considerably over the past two decades. The traditional view towards network analysis is limited to modelling interactions among entities of interest, for instance social networks or world wide web, and learning algorithms based on graph theory have been commonly used to solve these problems (Von Luxburg, 2007; Yan et al., 2006).