Understanding the Loss Surface of Neural Networks for Binary Classification
Preprint
Abstract
It is widely conjectured that the reason that training algorithms for neural networks are successful because all local minima lead to similar performance, for example, see (LeCun et al., 2015, Choromanska et al., 2015, Dauphin et al., 2014). Performance is typically measured in terms of two metrics: training performance and generalization performance. Here we focus on the training performance of single-layered neural networks for binary classification, and provide conditions under which the training error is zero at all local minima of a smooth hinge loss function. Our conditions are roughly in the following form: the neurons have to be strictly convex and the surrogate loss function should be a smooth version of hinge loss. We also provide counterexamples to show that when the loss function is replaced with quadratic loss or logistic loss, the result may not hold.Keywords
All Related Versions
This publication has 6 references indexed in Scilit:
- Theoretical Insights Into the Optimization Landscape of Over-Parameterized Shallow Neural NetworksIEEE Transactions on Information Theory, 2018
- ImageNet classification with deep convolutional neural networksCommunications of the ACM, 2017
- Deep Residual Learning for Image Recognition2016
- Sparse PCA. Extracting multi-scale structure from data2002
- Eigenfaces vs. Fisherfaces: recognition using class specific linear projectionIEEE Transactions on Pattern Analysis and Machine Intelligence, 1997
- Neural networks and principal component analysis: Learning from examples without local minimaNeural Networks, 1989