The Early Phase of Neural Network Training

April 27, 2020


Recent studies have shown that many important aspects of neural network learning take place within the very earliest iterations or epochs of training. For example, sparse, trainable sub-networks emerge (Frankle et al., 2019), gradient descent moves into a small subspace (Gur-Ari et al., 2018), and the network undergoes a critical period (Achille et al., 2019). Here we examine the changes that deep neural networks undergo during this early phase of training. We perform extensive measurements of the network state during these early iterations of training and leverage the framework of Frankle et al. (2019) to quantitatively probe the weight distribution and its reliance on various aspects of the dataset. We find that, within this framework, deep networks are not robust to reinitializing with random weights while maintaining signs, and that weight distributions are highly non-independent even after only a few hundred iterations. Despite this behavior, pre-training with blurred inputs or an auxiliary self-supervised task can approximate the changes in supervised networks, suggesting that these changes are not inherently label-dependent, though labels significantly accelerate this process. Together, these results help to elucidate the network changes occurring during this pivotal initial period of learning.

Download the Paper


Written by

Jonathan Frankle

David J. Schwab

Ari Morcos

Recent Publications

January 01, 2021

Asynchronous Gradient-Push | Facebook AI Research

We consider a multi-agent framework for distributed optimization where each agent has access to a local smooth strongly convex function, and the collective goal is to achieve consensus on the parameters that minimize the sum of the agents’…

Mahmoud Assran, Michael Rabbat

January 01, 2021

August 22, 2020

GrokNet: Unified Computer Vision Model Trunk and Embeddings For Commerce

In this paper we propose image classification modeling technique targeted for marketplace. We use public posts from marketplace and search log interactions for training image classifier and achieve significant improvements in e-commerce in comparison to previous version of our image classifier.

Sean Bell, Yiqun Liu, Sami Alsheikh, Yina Tang, Ed Pizzi, M. Henning, Karun Singh, Omkar Parkhi, Fedor Borisyuk

August 22, 2020

June 16, 2020


PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization

Due to memory limitations in current hardware, previous approaches tend to take low resolution images as input to cover large spatial context, and produce less precise (or low resolution) 3D estimates as a result. We address this limitation by formulating a multi-level architecture that is end-to-end trainable

Shunsuke Saito, Tomas Simon, Jason Saragih, Hanbyul Joo

June 16, 2020

June 14, 2020

Iterative Answer Prediction with Pointer-Augmented Multimodal Transformers for TextVQA | Facebook AI Research

Many visual scenes contain text that carries crucial information, and it is thus essential to understand text in images for downstream reasoning tasks. For example, a deep water label on a warning sign warns people about the danger in the…

Ronghang Hu, Amanpreet Singh, Trevor Darrell, Marcus Rohrbach

June 14, 2020

Help Us Pioneer The Future of AI

We share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.