Paper for June 29, 2020

[cbaziotis]

Suggest a paper you would like us to discuss in our next meeting. You can upvote a paper using the 👍 emoji.
The paper with the most upvotes by the end of the week (Friday) will be chosen.

[cbaziotis]

SynFlow: Pruning neural networks without any data by iteratively conserving synaptic flow https://arxiv.org/abs/2006.05467

Pruning the parameters of deep neural networks has generated intense interest due to potential savings in time, memory and energy both during training and at test time. Recent works have identified, through an expensive sequence of training and pruning cycles, the existence of winning lottery tickets or sparse trainable subnetworks at initialization. This raises a foundational question: can we identify highly sparse trainable subnetworks at initialization, without ever training, or indeed without ever looking at the data? We provide an affirmative answer to this question through theory driven algorithm design. We first mathematically formulate and experimentally verify a conservation law that explains why existing gradient-based pruning algorithms at initialization suffer from layer-collapse, the premature pruning of an entire layer rendering a network untrainable. This theory also elucidates how layer-collapse can be entirely avoided, motivating a novel pruning algorithm Iterative Synaptic Flow Pruning (SynFlow). This algorithm can be interpreted as preserving the total flow of synaptic strengths through the network at initialization subject to a sparsity constraint. Notably, this algorithm makes no reference to the training data and consistently outperforms existing state-of-the-art pruning algorithms at initialization over a range of models (VGG and ResNet), datasets (CIFAR-10/100 and Tiny ImageNet), and sparsity constraints (up to 99.9 percent). Thus our data-agnostic pruning algorithm challenges the existing paradigm that data must be used to quantify which synapses are important.

Video Review: https://www.youtube.com/watch?v=8l-TDqpoUQs

[cbaziotis]

Deep Differential System Stability -- Learning advanced computations from examples https://arxiv.org/abs/2006.06462

Can advanced mathematical computations be learned from examples? Using transformers over large generated datasets, we train models to learn properties of differential systems, such as local stability, behavior at infinity and controllability. We achieve near perfect estimates of qualitative characteristics of the systems, and good approximations of numerical quantities, demonstrating that neural networks can learn advanced theorems and complex computations without built-in mathematical knowledge.

Video review: https://www.youtube.com/watch?v=l12GXD0t_RE

[cbaziotis]

Prototypical Contrastive Learning of Unsupervised Representations
https://arxiv.org/abs/2005.04966
This paper presents Prototypical Contrastive Learning (PCL), an unsupervised representation learning method that addresses the fundamental limitations of instance-wise contrastive learning. PCL not only learns low-level features for the task of instance discrimination, but more importantly, it implicitly encodes semantic structures of the data into the learned embedding space. Specifically, we introduce prototypes as latent variables to help find the maximum-likelihood estimation of the network parameters in an Expectation-Maximization framework. We iteratively perform E-step as finding the distribution of prototypes via clustering and M-step as optimizing the network via contrastive learning. We propose ProtoNCE loss, a generalized version of the InfoNCE loss for contrastive learning, which encourages representations to be closer to their assigned prototypes. PCL achieves state-of-the-art results on multiple unsupervised representation learning benchmarks, with >10% accuracy improvement in low-resource transfer tasks.

Blog post: https://blog.einstein.ai/prototypical-contrastive-learning-pushing-the-frontiers-of-unsupervised-learning/

[cbaziotis]

A Probabilistic Formulation of Unsupervised Text Style Transfer
https://openreview.net/forum?id=HJlA0C4tPS

Video: https://iclr.cc/virtual_2020/poster_HJlA0C4tPS.html

TL;DR: We formulate a probabilistic latent sequence model to tackle unsupervised text style transfer, and show its effectiveness across a suite of unsupervised text style transfer tasks.
Abstract: We present a deep generative model for unsupervised text style transfer that unifies previously proposed non-generative techniques. Our probabilistic approach models non-parallel data from two domains as a partially observed parallel corpus. By hypothesizing a parallel latent sequence that generates each observed sequence, our model learns to transform sequences from one domain to another in a completely unsupervised fashion. In contrast with traditional generative sequence models (e.g. the HMM), our model makes few assumptions about the data it generates: it uses a recurrent language model as a prior and an encoder-decoder as a transduction distribution. While computation of marginal data likelihood is intractable in this model class, we show that amortized variational inference admits a practical surrogate. Further, by drawing connections between our variational objective and other recent unsupervised style transfer and machine translation techniques, we show how our probabilistic view can unify some known non-generative objectives such as backtranslation and adversarial loss. Finally, we demonstrate the effectiveness of our method on a wide range of unsupervised style transfer tasks, including sentiment transfer, formality transfer, word decipherment, author imitation, and related language translation. Across all style transfer tasks, our approach yields substantial gains over state-of-the-art non-generative baselines, including the state-of-the-art unsupervised machine translation techniques that our approach generalizes. Further, we conduct experiments on a standard unsupervised machine translation task and find that our unified approach matches the current state-of-the-art.
Keywords: unsupervised text style transfer, deep latent sequence model

[cbaziotis]

How Context Affects Language Models' Factual Predictions
https://openreview.net/forum?id=025X0zPfn

Abstract: When pre-trained on large unsupervised textual corpora, language models are able to store and retrieve factual knowledge to some extent, making it possible to use them directly for zero-shot cloze-style question answering. However, storing factual knowledge in a fixed number of weights of a language model clearly has limitations. Previous approaches have successfully provided access to information outside the model weights using supervised architectures that combine an information retrieval system with a machine reading component. In this paper, we go one step further and integrate information from a retrieval system with a pre-trained language model in a purely unsupervised way. We report that augmenting pre-trained language models in this way dramatically improves performance and that it is competitive with a supervised machine reading baseline without requiring any supervised training. Furthermore, processing query and context with different segment tokens allows BERT to utilize its Next Sentence Prediction pre-trained classifier to determine whether the context is relevant or not, substantially improving BERT's zero-shot cloze-style question-answering performance and making its predictions robust to noisy contexts.

Comment: This paper studies how factual predictions of a Masked Language Model (MLM) are influenced by appending additional context via various context construction methods. The work presents a set of interesting probes for the analysis, with good justification on the probe design. The paper is well written, clear, and provides good insights on understanding and improving MLM.