Exploring the Limits of Large Scale Pre-training
Recent developments in large-scale machine learning suggest that by scaling up data, model size and training time properly, one might observe that improvements in pre-training would transfer favorably to most downstream tasks. In this work, we systematically study this phenomena and establish that, as we increase the upstream accuracy, the performance of downstream tasks saturates. In particular, we investigate more than 4800 experiments on Vision Transformers, MLP-Mixers and ResNets with number of parameters ranging from ten million to ten billion, trained on the largest scale of available image data (JFT, ImageNet21K) and evaluated on more than 20 downstream image recognition tasks. We propose a model for downstream performance that reflects the saturation phenomena and captures the nonlinear relationship in performance of upstream and downstream tasks. Delving deeper to understand the reasons that give rise to these phenomena, we show that the saturation behavior we observe is closely related to the way that representations evolve through the layers of the models. We showcase an even more extreme scenario where performance on upstream and downstream are at odds with each other. That is, to have a better downstream performance, we need to hurt upstream accuracy.
Hanie Sedghi is a Research Scientist at Allen Institute for Artificial Intelligence (AI2). Her research interests include large-scale machine learning, high-dimensional statistics and probabilistic models. More recently, she has been working on inference and learning in latent variable models. She has received her Ph.D. from University of Southern California with a minor in Mathematics in 2015. She was also a visiting researcher at University of California, Irvine working with professor Anandkumar during her Ph.D. She received her B.Sc. and M.Sc. degree from Sharif University of Technology, Tehran, Iran.