EHI: End-to-end Learning of Hierarchical Index for Efficient Dense Retrieval

Dense embedding-based retrieval is now the industry standard for semantic search and ranking problems, like obtaining relevant web documents for a given query. Such techniques use a two-stage process: (a) contrastive learning to train a dual encoder to embed both the query and documents and (b) approximate nearest neighbor search (ANNS) for finding similar documents for a given query. These two stages are disjoint; the learned embeddings might be ill-suited for the ANNS method and vice-versa, leading to suboptimal performance. In this work, we propose End-to-end Hierarchical Indexing -- EHI -- that jointly learns both the embeddings and the ANNS structure to optimize retrieval performance. EHI uses a standard dual encoder model for embedding queries and documents while learning an inverted file index (IVF) style tree structure for efficient ANNS. To ensure stable and efficient learning of discrete tree-based ANNS structure, EHI introduces the notion of dense path embedding that captures the position of a query/document in the tree. We demonstrate the effectiveness of EHI on several benchmarks, including de-facto industry standard MS MARCO (Dev set and TREC DL19) datasets. For example, with the same compute budget, EHI outperforms state-of-the-art (SOTA) in by 0.6% (MRR@10) on MS MARCO dev set and by 4.2% (nDCG@10) on TREC DL19 benchmarks.

Stochastic Re-weighted Gradient Descent via Distributionally Robust Optimization

We develop a re-weighted gradient descent technique for boosting the performance of deep neural networks. Our algorithm involves the importance weighting of data points during each optimization step. Our approach is inspired by distributionally robust optimization with $f$-divergences, which has been known to result in models with improved generalization guarantees. Our re-weighting scheme is simple, computationally efficient, and can be combined with any popular optimization algorithms such as SGD and Adam. Empirically, we demonstrate our approach's superiority on various tasks, including vanilla classification, classification with label imbalance, noisy labels, domain adaptation, and tabular representation learning. Notably, we obtain improvements of +0.7% and +1.44% over SOTA on DomainBed and Tabular benchmarks, respectively. Moreover, our algorithm boosts the performance of BERT on GLUE benchmarks by +1.94%, and ViT on ImageNet-1K by +0.9%. These results demonstrate the effectiveness of the proposed approach, indicating its potential for improving performance in diverse domains.

Look Back When Surprised: Stabilizing Reverse Experience Replay for Neural Approximation

Experience replay methods, which are an essential part of reinforcement learning(RL) algorithms, are designed to mitigate spurious correlations and biases while learning from temporally dependent data. Roughly speaking, these methods allow us to draw batched data from a large buffer such that these temporal correlations do not hinder the performance of descent algorithms. In this experimental work, we consider the recently developed and theoretically rigorous reverse experience replay (RER), which has been shown to remove such spurious biases in simplified theoretical settings. We combine RER with optimistic experience replay (OER) to obtain RER++, which is stable under neural function approximation. We show via experiments that this has a better performance than techniques like prioritized experience replay (PER) on various tasks, with a significantly smaller computational complexity. It is well known in the RL literature that choosing examples greedily with the largest TD error (as in OER) or forming mini-batches with consecutive data points (as in RER) leads to poor performance. However, our method, which combines these techniques, works very well.

Rethinking Learning Dynamics in RL using Adversarial Networks

We present a learning mechanism for reinforcement learning of closely related skills parameterized via a skill embedding space. Our approach is grounded on the intuition that nothing makes you learn better than a coevolving adversary. The main contribution of our work is to formulate an adversarial training regime for reinforcement learning with the help of entropy-regularized policy gradient formulation. We also adapt existing measures of causal attribution to draw insights from the skills learned. Our experiments demonstrate that the adversarial process leads to a better exploration of multiple solutions and understanding the minimum number of different skills necessary to solve a given set of tasks.

The Effect of Diversity in Meta-Learning

Few-shot learning aims to learn representations that can tackle novel tasks given a small number of examples. Recent studies show that task distribution plays a vital role in the model's performance. Conventional wisdom is that task diversity should improve the performance of meta-learning. In this work, we find evidence to the contrary; we study different task distributions on a myriad of models and datasets to evaluate the effect of task diversity on meta-learning algorithms. For this experiment, we train on multiple datasets, and with three broad classes of meta-learning models - Metric-based (i.e., Protonet, Matching Networks), Optimization-based (i.e., MAML, Reptile, and MetaOptNet), and Bayesian meta-learning models (i.e., CNAPs). Our experiments demonstrate that the effect of task diversity on all these algorithms follows a similar trend, and task diversity does not seem to offer any benefits to the learning of the model. Furthermore, we also demonstrate that even a handful of tasks, repeated over multiple batches, would be sufficient to achieve a performance similar to uniform sampling and draws into question the need for additional tasks to create better models.