Effect of Initial Configuration of Weights on Training and Function of Artificial Neural Networks
Open Access
- 13 September 2021
- journal article
- research article
- Published by MDPI AG in Mathematics
- Vol. 9 (18) , 2246
Abstract
The function and performance of neural networks are largely determined by the evolution of their weights and biases in the process of training, starting from the initial configuration of these parameters to one of the local minima of the loss function. We perform the quantitative statistical characterization of the deviation of the weights of two-hidden-layer feedforward ReLU networks of various sizes trained via Stochastic Gradient Descent (SGD) from their initial random configuration. We compare the evolution of the distribution function of this deviation with the evolution of the loss during training. We observed that successful training via SGD leaves the network in the close neighborhood of the initial configuration of its weights. For each initial weight of a link we measured the distribution function of the deviation from this value after training and found how the moments of this distribution and its peak depend on the initial weight. We explored the evolution of these deviations during training and observed an abrupt increase within the overfitting region. This jump occurs simultaneously with a similarly abrupt increase recorded in the evolution of the loss function. Our results suggest that SGD’s ability to efficiently find local minima is restricted to the vicinity of the random initial configuration of weights.Keywords
Funding Information
- Fundação para a Ciência e a Tecnologia (UIDP/50025/2020, UIDB/50025/2020, CEECIND/04697/2017, UIDB/50008/2020-UIDP/50008/2020)
This publication has 10 references indexed in Scilit:
- Wide neural networks of any depth evolve as linear models under gradient descent *Journal of Statistical Mechanics: Theory and Experiment, 2020
- What’s Hidden in a Randomly Weighted Neural Network?2020
- Toward Moderate Overparameterization: Global Convergence Guarantees for Training Shallow Neural NetworksIEEE Journal on Selected Areas in Information Theory, 2020
- Gradient descent optimizes over-parameterized deep ReLU networksMachine Learning, 2019
- Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification2015
- Deep learningNature, 2015
- The relative age effect in youth soccer across EuropeJournal of Sports Sciences, 2005
- A weight initialization method for improving training speed in feedforward neural networkNeurocomputing, 2000
- Efficient BackProp1998
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998