This paper uses information-theoretic tools to analyze the generalization error in unsupervised domain adaptation (UDA). The first notion measures the gap between the population risk in the target domain and that in the source domain, and the second measures the gap between the population risk in the target domain and the empirical risk in the source domain. While our bounds for the first kind of error are in line with the traditional analysis and give similar insights, our bounds on the second kind of error are algorithm-dependent, which also provide insights into algorithm designs. This paper focuses on the unsupervised domain adaptation (UDA) task, where the learner is confronted with a source domain and a target domain and the algorithm is allowed to access to a labeled training sample from the source domain and an unlabeled training sample from the target domain. The goal is to find a predictor that performs well on the target domain. A main obstacle in such a task is the discrepancy between the two domains. For example, Nguyen et al. (2022) uses a (reverse) KL divergence to measure the misalignment of the two domain distributions, and motivated by their generalization bound, they design an algorithm that penalizes the KL divergence between the marginal distributions of two domains in the representation space. Despite that this "KL guided domain adaptation" algorithm is demonstrated to outperform many existing marginal alignment algorithms (Ganin et al., 2016; Sun & Saenko, 2016; Shen et al., 2018; Li et al., 2018), it is not clear whether KL-based alignment of marginal distributions is adequate for UDA, and more fundamentally, what role the unlabelled target-domain sample should play in cross-domain generalization. Notably, most UDA algorithms are heuristically designed and intuitively justified. Moreover, most existing generalization bounds are algorithm-independent.

While graph neural networks (GNNs) have been successful for node classification tasks and link prediction tasks in graph, learning graph-level representations still remains a challenge. For the graph-level representation, it is important to learn both representation of neighboring nodes, i.e., aggregation, and graph structural information. A number of graph pooling methods have been developed for this goal. However, most of the existing pooling methods utilize k-hop neighborhood without considering explicit structural information in a graph. In this paper, we propose Structure Prototype Guided Pooling (SPGP) that utilizes prior graph structures to overcome the limitation. SPGP formulates graph structures as learnable prototype vectors and computes the affinity between nodes and prototype vectors. This leads to a novel node scoring scheme that prioritizes informative nodes while encapsulating the useful structures of the graph. Our experimental results show that SPGP outperforms state-of-the-art graph pooling methods on graph classification benchmark datasets in both accuracy and scalability.

Transformers have become a default architecture in computer vision, but understanding what drives their predictions remains a challenging problem. Current explanation approaches rely on attention values or input gradients, but these provide a limited view of a model's dependencies. Shapley values offer a theoretically sound alternative, but their computational cost makes them impractical for large, high-dimensional models. In this work, we aim to make Shapley values practical for vision transformers (ViTs). To do so, we first leverage an attention masking approach to evaluate ViTs with partial information, and we then develop a procedure to generate Shapley value explanations via a separate, learned explainer model. Our experiments compare Shapley values to many baseline methods (e.g., attention rollout, GradCAM, LRP), and we find that our approach provides more accurate explanations than existing methods for ViTs. Transformers (Vaswani et al., 2017) were originally introduced for ...

The success of deep learning is due in large part to our ability to solve certain massive non-convex optimization problems with relative ease. Though non-convex optimization is NP-hard, simple algorithms -- often variants of stochastic gradient descent -- exhibit surprising effectiveness in fitting large neural networks in practice. We argue that neural network loss landscapes often contain (nearly) a single basin after accounting for all possible permutation symmetries of hidden units a la Entezari et al. 2021. We introduce three algorithms to permute the units of one model to bring them into alignment with a reference model in order to merge the two models in weight space. This transformation produces a functionally equivalent set of weights that lie in an approximately convex basin near the reference model. Experimentally, we demonstrate the single basin phenomenon across a variety of model architectures and datasets, including the first (to our knowledge) demonstration of zero-barrier linear mode connectivity between independently trained ResNet models on CIFAR-10. Additionally, we identify intriguing phenomena relating model width and training time to mode connectivity. Finally, we discuss shortcomings of the linear mode connectivity hypothesis, including a counterexample to the single basin theory.

We analyze the complexity of sampling from a class of heavy-tailed distributions by discretizing a natural class of It\^o diffusions associated with weighted Poincar\'e inequalities. Based on a mean-square analysis, we establish the iteration complexity for obtaining a sample whose distribution is $\epsilon$ close to the target distribution in the Wasserstein-2 metric. In this paper, our results take the mean-square analysis to its limits, i.e., we invariably only require that the target density has finite variance, the minimal requirement for a mean-square analysis. To obtain explicit estimates, we compute upper bounds on certain moments associated with heavy-tailed targets under various assumptions. We also provide similar iteration complexity results for the case where only function evaluations of the unnormalized target density are available by estimating the gradients using a Gaussian smoothing technique. We provide illustrative examples based on the multivariate $t$-distribution.

Imagine that you are a Data Scientist working for a retail company and your boss requests for the customers' segmentation into the following groups: low, average, medium, or platinum customers based on spending behavior for targeted marketing purposes and product recommendations. Knowing that there is no such historical label associated with those customers, how is it possible to categorize them? This is where clustering can help. It is an unsupervised machine-learning technique used to group unlabeled data into similar categories or clusters. This conceptual article will focus more on the K-means clustering approach, one of the many techniques in unsupervised machine learning.

Abstract: We investigate the general properties of the dimensional reduction of the Dirac theory, formulated in a Minkowski spacetime with an arbitrary number of spatial dimensions. This is done by applying Hadamard's method of descent, which consists in conceiving low-dimensional theories as a specialization of high-dimensional ones that are uniform along the additional space coordinate. We show that the Dirac equation reduces to either a single Dirac equation or two decoupled Dirac equations, depending on whether the higher-dimensional manifold has even or odd spatial dimensions, respectively. Abstract: Sparse reduced rank regression is an essential statistical learning method. In the contemporary literature, estimation is typically formulated as a nonconvex optimization that often yields to a local optimum in numerical computation. Yet, their theoretical analysis is always centered on the global optimum, resulting in a discrepancy between the statistical guarantee and the numerical computation.

We develop an efficient machine learning protocol to predict the noise-induced coherence from the nonequilibrium fluctuations of photon exchange statistics in a quantum heat engine. The engine is a four-level quantum system coupled to a unimodal quantum cavity. The nonequilibrium fluctuations correspond to the work done during the photon exchange process between the four-level system and the cavity mode. We specifically evaluate the mean, variance, skewness, and kurtosis for a range of engine parameters using a full counting statistical approach combined with a quantum master equation technique. We use these numerically evaluated cumulants as input data to successfully predict the hot bath induced coherence. A supervised machine learning technique based on K-Nearest Neighbor(KNN) is found to work better than a variety of learning models that we tested.

Molecular-level understanding of the interactions between the constituents of an atomic structure is essential for designing novel materials in various applications. This need goes beyond the basic knowledge of the number and types of atoms, their chemical composition, and the character of the chemical interactions. The bigger picture takes place on the quantum level which can be addressed by using the Density-functional theory (DFT). Use of DFT, however, is a computationally taxing process, and its results do not readily provide easily interpretable insight into the atomic interactions which would be useful information in material design. An alternative way to address atomic interactions is to use an interpretable machine learning approach, where a predictive DFT surrogate is constructed and analyzed. The purpose of this paper is to propose such a procedure using a modification of the recently published interpretable distance-based regression method. Our tests with a representative benchmark set of molecules and a complex hybrid nanoparticle confirm the viability and usefulness of the proposed approach.

In Multi-Label Learning (MLL), it is extremely challenging to accurately annotate every appearing object due to expensive costs and limited knowledge. When facing such a challenge, a more practical and cheaper alternative should be Single Positive Multi-Label Learning (SPMLL), where only one positive label needs to be provided per sample. Existing SPMLL methods usually assume unknown labels as negatives, which inevitably introduces false negatives as noisy labels. More seriously, Binary Cross Entropy (BCE) loss is often used for training, which is notoriously not robust to noisy labels. To mitigate this issue, we customize an objective function for SPMLL by pushing only one pair of labels apart each time to prevent the domination of negative labels, which is the main culprit of fitting noisy labels in SPMLL. To further combat such noisy labels, we explore the high-rankness of label matrix, which can also push apart different labels. By directly extending from SPMLL to MLL with full labels, a unified loss applicable to both settings is derived. Experiments on real datasets demonstrate that the proposed loss not only performs more robustly to noisy labels for SPMLL but also works well for full labels. Besides, we empirically discover that high-rankness can mitigate the dramatic performance drop in SPMLL. Most surprisingly, even without any regularization or fine-tuned label correction, only adopting our loss defeats state-of-the-art SPMLL methods on CUB, a dataset that severely lacks labels.