An Optimal Transport Framework for Zero-Shot Learning

We present an optimal transport (OT) framework for generalized zero-shot learning (GZSL) of imaging data, seeking to distinguish samples for both seen and unseen classes, with the help of auxiliary attributes. The discrepancy between features and attributes is minimized by solving an optimal transport problem. {Specifically, we build a conditional generative model to generate features from seen-class attributes, and establish an optimal transport between the distribution of the generated features and that of the real features.} The generative model and the optimal transport are optimized iteratively with an attribute-based regularizer, that further enhances the discriminative power of the generated features. A classifier is learned based on the features generated for both the seen and unseen classes. In addition to generalized zero-shot learning, our framework is also applicable to standard and transductive ZSL problems. Experiments show that our optimal transport-based method outperforms state-of-the-art methods on several benchmark datasets.

Gaussian-Process-Based Dynamic Embedding for Textual Networks

Textual network embedding aims to learn low-dimensional representations of text-annotated nodes in a graph. Prior works have typically focused on fixed graph structures. However, real-world networks are often dynamic. We address this challenge with a novel end-to-end node-embedding model, called Dynamic Embedding for Textual Networks with a Gaussian Process (DetGP). Because the structure is allowed to be dynamic, our method uses the Gaussian process to take advantage of its non-parametric properties. After training, DetGP can be applied efficiently to dynamic graphs without re-training or backpropagation. To use both local and global graph structures, diffusion is used to model multiple hops between neighbors. The relative importance of global versus local structure for the embeddings is learned automatically. With the non-parametric nature of the Gaussian process, updating the embeddings for a changed graph structure requires only a forward pass through the learned model. Experiments demonstrate the empirical effectiveness of our method compared to baseline approaches, on link prediction and node classification. We further show that DetGP can be straightforwardly and efficiently applied to dynamic textual networks.

Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods

We investigate time-dependent data analysis from the perspective of recurrent kernel machines, from which models with hidden units and gated memory cells arise naturally. By considering dynamic gating of the memory cell, a model closely related to the long short-term memory (LSTM) recurrent neural network is derived. Extending this setup to $n$-gram filters, the convolutional neural network (CNN), Gated CNN, and recurrent additive network (RAN) are also recovered as special cases. Our analysis provides a new perspective on the LSTM, while also extending it to $n$-gram convolutional filters. Experiments are performed on natural language processing tasks and on analysis of local field potentials (neuroscience). We demonstrate that the variants we derive from kernels perform on par or even better than traditional neural methods. For the neuroscience application, the new models demonstrate significant improvements relative to the prior state of the art.

Straight-Through Estimator as Projected Wasserstein Gradient Flow

The Straight-Through (ST) estimator is a widely used technique for back-propagating gradients through discrete random variables. However, this effective method lacks theoretical justification. In this paper, we show that ST can be interpreted as the simulation of the projected Wasserstein gradient flow (pWGF). Based on this understanding, a theoretical foundation is established to justify the convergence properties of ST. Further, another pWGF estimator variant is proposed, which exhibits superior performance on distributions with infinite support,e.g., Poisson distributions. Empirically, we show that ST and our proposed estimator, while applied to different types of discrete structures (including both Bernoulli and Poisson latent variables), exhibit comparable or even better performances relative to other state-of-the-art methods. Our results uncover the origin of the widespread adoption of the ST estimator and represent a helpful step towards exploring alternative gradient estimators for discrete variables.

Fused Gromov-Wasserstein Alignment for Hawkes Processes

We propose a novel fused Gromov-Wasserstein alignment method to jointly learn the Hawkes processes in different event spaces, and align their event types. Given two Hawkes processes, we use fused Gromov-Wasserstein discrepancy to measure their dissimilarity, which considers both the Wasserstein discrepancy based on their base intensities and the Gromov-Wasserstein discrepancy based on their infectivity matrices. Accordingly, the learned optimal transport reflects the correspondence between the event types of these two Hawkes processes. The Hawkes processes and their optimal transport are learned jointly via maximum likelihood estimation, with a fused Gromov-Wasserstein regularizer. Experimental results show that the proposed method works well on synthetic and real-world data.

Improving Textual Network Learning with Variational Homophilic Embeddings

The performance of many network learning applications crucially hinges on the success of network embedding algorithms, which aim to encode rich network information into low-dimensional vertex-based vector representations. This paper considers a novel variational formulation of network embeddings, with special focus on textual networks. Different from most existing methods that optimize a discriminative objective, we introduce Variational Homophilic Embedding (VHE), a fully generative model that learns network embeddings by modeling the semantic (textual) information with a variational autoencoder, while accounting for the structural (topology) information through a novel homophilic prior design. Homophilic vertex embeddings encourage similar embedding vectors for related (connected) vertices. The proposed VHE promises better generalization for downstream tasks, robustness to incomplete observations, and the ability to generalize to unseen vertices. Extensive experiments on real-world networks, for multiple tasks, demonstrate that the proposed method consistently achieves superior performance relative to competing state-of-the-art approaches.

Ouroboros: On Accelerating Training of Transformer-Based Language Models

Language models are essential for natural language processing (NLP) tasks, such as machine translation and text summarization. Remarkable performance has been demonstrated recently across many NLP domains via a Transformer-based language model with over a billion parameters, verifying the benefits of model size. Model parallelism is required if a model is too large to fit in a single computing device. Current methods for model parallelism either suffer from backward locking in backpropagation or are not applicable to language models. We propose the first model-parallel algorithm that speeds the training of Transformer-based language models. We also prove that our proposed algorithm is guaranteed to converge to critical points for non-convex problems. Extensive experiments on Transformer and Transformer-XL language models demonstrate that the proposed algorithm obtains a much faster speedup beyond data parallelism, with comparable or better accuracy. Code to reproduce experiments is to be found at \url{https://github.com/LaraQianYang/Ouroboros}.

Contrastively Smoothed Class Alignment for Unsupervised Domain Adaptation

Recent unsupervised approaches to domain adaptation primarily focus on minimizing the gap between the source and the target domains through refining the feature generator, in order to learn a better alignment between the two domains. This minimization can be achieved via a domain classifier to detect target-domain features that are divergent from source-domain features. However, by optimizing via such domain classification discrepancy, ambiguous target samples that are not smoothly distributed on the low-dimensional data manifold are often missed. To solve this issue, we propose a novel Contrastively Smoothed Class Alignment (CoSCA) model, that explicitly incorporates both intra- and inter-class domain discrepancy to better align ambiguous target samples with the source domain. CoSCA estimates the underlying label hypothesis of target samples, and simultaneously adapts their feature representations by optimizing a proposed contrastive loss. In addition, Maximum Mean Discrepancy (MMD) is utilized to directly match features between source and target samples for better global alignment. Experiments on several benchmark datasets demonstrate that CoSCA can outperform state-of-the-art approaches for unsupervised domain adaptation by producing more discriminative features.

LMVP: Video Predictor with Leaked Motion Information

We propose a Leaked Motion Video Predictor (LMVP) to predict future frames by capturing the spatial and temporal dependencies from given inputs. The motion is modeled by a newly proposed component, motion guider, which plays the role of both learner and teacher. Specifically, it {\em learns} the temporal features from real data and {\em guides} the generator to predict future frames. The spatial consistency in video is modeled by an adaptive filtering network. To further ensure the spatio-temporal consistency of the prediction, a discriminator is also adopted to distinguish the real and generated frames. Further, the discriminator leaks information to the motion guider and the generator to help the learning of motion. The proposed LMVP can effectively learn the static and temporal features in videos without the need for human labeling. Experiments on synthetic and real data demonstrate that LMVP can yield state-of-the-art results.