A reliable and efficient representation of multivariate time series is crucial in various downstream machine learning tasks. In multivariate time series forecasting, each variable depends on its historical values and there are inter-dependencies among variables as well. Models have to be designed to capture both intra- and inter-relationships among the time series. To move towards this goal, we propose the Time Series Attention Transformer (TSAT) for multivariate time series representation learning. Using TSAT, we represent both temporal information and inter-dependencies of multivariate time series in terms of edge-enhanced dynamic graphs. The intra-series correlations are represented by nodes in a dynamic graph; a self-attention mechanism is modified to capture the inter-series correlations by using the super-empirical mode decomposition (SMD) module. We applied the embedded dynamic graphs to times series forecasting problems, including two real-world datasets and two benchmark datasets. Extensive experiments show that TSAT clearly outerperforms six state-of-the-art baseline methods in various forecasting horizons. We further visualize the embedded dynamic graphs to illustrate the graph representation power of TSAT. We share our code at https://github.com/RadiantResearch/TSAT.

Robust speech emotion recognition relies on the quality of the speech features. We present speech features enhancement strategy that improves speech emotion recognition. We used the INTERSPEECH 2010 challenge feature-set. We identified subsets from the features set and applied Principle Component Analysis to the subsets. Finally, the features are fused horizontally. The resulting feature set is analyzed using t-distributed neighbour embeddings (t-SNE) before the application of features for emotion recognition. The method is compared with the state-of-the-art methods used in the literature. The empirical evidence is drawn using two well-known datasets: Emotional Speech Dataset (EMO-DB) and Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) for two languages, German and English, respectively. Our method achieved an average recognition gain of 11.5\% for six out of seven emotions for the EMO-DB dataset, and 13.8\% for seven out of eight emotions for the RAVDESS dataset as compared to the baseline study.

Graph embedding techniques are a staple of modern graph learning research. When using embeddings for downstream tasks such as classification, information about their stability and robustness, i.e., their susceptibility to sources of noise, stochastic effects, or specific parameter choices, becomes increasingly important. As one of the most prominent graph embedding schemes, we focus on node2vec and analyse its embedding quality from multiple perspectives. Our findings indicate that embedding quality is unstable with respect to parameter choices, and we propose strategies to remedy this in practice.

Modern machine learning models use an ever-increasing number of parameters to train (175 billion parameters for GPT-3) with large datasets to obtain better performance. Bigger is better has been the norm. Optical computing has been reawakened as a potential solution to large-scale computing through optical accelerators that carry out linear operations while reducing electrical power. However, to achieve efficient computing with light, creating and controlling nonlinearity optically rather than electronically remains a challenge. This study explores a reservoir computing (RC) approach whereby a 14 mm long few-mode waveguide in LiNbO3 on insulator is used as a complex nonlinear optical processor. A dataset is encoded digitally on the spectrum of a femtosecond pulse which is then launched in the waveguide. The output spectrum depends nonlinearly on the input. We experimentally show that a simple digital linear classifier with 784 parameters using the output spectrum from the waveguide as input increased the classification accuracy of several databases compared to non-transformed data, approximately 10$\%$. In comparison, a deep digital neural network (NN) with 40000 parameters was necessary to achieve the same accuracy. Reducing the number of parameters by a factor of $\sim$50 illustrates that a compact optical RC approach can perform on par with a deep digital NN.

Generalization analyses of deep learning typically assume that the training converges to a fixed point. But, recent results indicate that in practice, the weights of deep neural networks optimized with stochastic gradient descent often oscillate indefinitely. To reduce this discrepancy between theory and practice, this paper focuses on the generalization of neural networks whose training dynamics do not necessarily converge to fixed points. Our main contribution is to propose a notion of statistical algorithmic stability (SAS) that extends classical algorithmic stability to non-convergent algorithms and to study its connection to generalization. This ergodic-theoretic approach leads to new insights when compared to the traditional optimization and learning theory perspectives. We prove that the stability of the time-asymptotic behavior of a learning algorithm relates to its generalization and empirically demonstrate how loss dynamics can provide clues to generalization performance. Our findings provide evidence that networks that "train stably generalize better" even when the training continues indefinitely and the weights do not converge.

Stackelberg games are a canonical model for strategic principal-agent interactions. Consider a defense system that distributes its security resources across high-risk targets prior to attacks being executed; or a tax policymaker who sets rules on when audits are triggered prior to seeing filed tax reports; or a seller who chooses a price prior to knowing a customer's proclivity to buy. In each of these scenarios, a principal first selects an action X and then an agent reacts with an action Y, where X and Y are the principal's and agent's action spaces, respectively. In the examples above, agent actions correspond to which target to attack, how much tax to pay to evade an audit, and how much to purchase, respectively. Typically, the principal wants an that maximizes their payoff when the agent plays a best response = br(); such a pair (,) is a Stackelberg equilibrium. By committing to a strategy, the principal can guarantee they achieve a higher payoff than in the fixed point equilibrium of the corresponding simultaneous-play game. However, finding such a strategy requires knowledge of the agent's payoff function. When faced with unknown agent payoffs, the principal can attempt to learn a best response via repeated interactions with the agent. If a (naïve) agent is unaware that such learning occurs and always plays a best response, the principal can use classical online learning approaches to optimize their own payoff in the stage game.

Natural language understanding (NLU) is the task of semantic decoding of human languages by machines. NLU models rely heavily on large training data to ensure good performance. However, substantial languages and domains have very few data resources and domain experts. It is necessary to overcome the data scarcity challenge, when very few or even zero training samples are available. In this thesis, we focus on developing cross-lingual and cross-domain methods to tackle the low-resource issues. First, we propose to improve the model's cross-lingual ability by focusing on the task-related keywords, enhancing the model's robustness and regularizing the representations. We find that the representations for low-resource languages can be easily and greatly improved by focusing on just the keywords. Second, we present Order-Reduced Modeling methods for the cross-lingual adaptation, and find that modeling partial word orders instead of the whole sequence can improve the robustness of the model against word order differences between languages and task knowledge transfer to low-resource languages. Third, we propose to leverage different levels of domain-related corpora and additional masking of data in the pre-training for the cross-domain adaptation, and discover that more challenging pre-training can better address the domain discrepancy issue in the task knowledge transfer. Finally, we introduce a coarse-to-fine framework, Coach, and a cross-lingual and cross-domain parsing framework, X2Parser. Coach decomposes the representation learning process into a coarse-grained and a fine-grained feature learning, and X2Parser simplifies the hierarchical task structures into flattened ones. We observe that simplifying task structures makes the representation learning more effective for low-resource languages and domains.

In many applications, machine learned (ML) models are required to hold some invariance qualities, such as rotation, size, intensity, and background invariance. Unlike many types of variance, the variants of background scenes cannot be ordered easily, which makes it difficult to analyze the robustness and biases of the models concerned. In this work, we present a technical solution for ordering background scenes according to their semantic proximity to a target image that contains a foreground object being tested. We make use of the results of object recognition as the semantic description of each image, and construct an ontology for storing knowledge about relationships among different objects using association analysis. This ontology enables (i) efficient and meaningful search for background scenes of different semantic distances to a target image, (ii) quantitative control of the distribution and sparsity of the sampled background scenes, and (iii) quality assurance using visual representations of invariance testing results (referred to as variance matrices). In this paper, we also report the training of an ML4ML assessor to evaluate the invariance quality of ML models automatically.

Predicting structural and energetic properties of a molecular system is one of the fundamental tasks in molecular simulations, and it has use cases in chemistry, biology, and medicine. In the past decade, the advent of machine learning algorithms has impacted on molecular simulations for various tasks, including property prediction of atomistic systems. In this paper, we propose a novel methodology for transferring knowledge obtained from simple molecular systems to a more complex one, possessing a significantly larger number of atoms and degrees of freedom. In particular, we focus on the classification of high and low free-energy states. Our approach relies on utilizing (i) a novel hypergraph representation of molecules, encoding all relevant information for characterizing the potential energy of a conformation, and (ii) novel message passing and pooling layers for processing and making predictions on such hypergraph-structured data. Despite the complexity of the problem, our results show a remarkable AUC of 0.92 for transfer learning from tri-alanine to the deca-alanine system. Moreover, we show that the very same transfer learning approach can be used to group, in an unsupervised way, various secondary structures of deca-alanine in clusters having similar free-energy values. Our study represents a proof of concept that reliable transfer learning models for molecular systems can be designed paving the way to unexplored routes in prediction of structural and energetic properties of biologically relevant systems.

In many problems in modern statistics and machine learning, it is often of interest to establish that a first order method on a non-convex risk function eventually enters a region of parameter space in which the risk is locally convex. We derive an asymptotic comparison inequality, which we call the Sudakov-Fernique post-AMP inequality, which, in a certain class of problems involving a GOE matrix, is able to probe properties of an optimization landscape locally around the iterates of an approximate message passing (AMP) algorithm. As an example of its use, we provide a new, and arguably simpler, proof of some of the results of Celentano et al. (2021), which establishes that the so-called TAP free energy in the $\mathbb{Z}_2$-synchronization problem is locally convex in the region to which AMP converges. We further prove a conjecture of El Alaoui et al. (2022) involving the local convexity of a related but distinct TAP free energy, which, as a consequence, confirms that their algorithm efficiently samples from the Sherrington-Kirkpatrick Gibbs measure throughout the "easy" regime.