Data-driven Weight Initialization with Sylvester Solvers
arXiv:2105.10335
Abstract
In this work, we propose a data-driven scheme to initialize the parameters of a deep neural network. This is in contrast to traditional approaches which randomly initialize parameters by sampling from transformed standard distributions. Such methods do not use the training data to produce a more informed initialization. Our method uses a sequential layer-wise approach where each layer is initialized using its input activations. The initialization is cast as an optimization problem where we minimize a combination of encoding and decoding losses of the input activations, which is further constrained by a user-defined latent code. The optimization problem is then restructured into the well-known Sylvester equation, which has fast and efficient gradient-free solutions. Our data-driven method achieves a boost in performance compared to random initialization methods, both before start of training and after training is over. We show that our proposed method is especially effective in few-shot and fine-tuning settings. We conclude this paper with analyses on time complexity and the effect of different latent codes on the recognition performance.
arXiv commentPractical Machine Learning for Developing Countries Workshop, International Conference on Learning Representations, 2021
References in corpus (7)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Language Models are Few-Shot Learners
- Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
- All you need is a good init
- Data-dependent Initializations of Convolutional Neural Networks
- Random Walk Initialization for Training Very Deep Feedforward Networks
- GradInit: Learning to Initialize Neural Networks for Stable and Efficient Training