Chapter 8 Contents
Overview
Section 8.1: Characteristic Features
Section 8.2: The Gradient is the Sum of Single-Pattern Gradients
Section 8.3: Weight-Space Symmetries
Section 8.4: Remarks
Section 8.5: Local Minima
Section 8.5.1: Single-Layer Nets Can Have Local Minima
Section 8.5.2: No Local Minima for Linearly Separable Data
Section 8.5.3: Local Minima Really Do Exist
Section 8.5.4: The Effect of Network Size
Section 8.6: Properties of the Hessian Matrix
Section 8.7: Gain Scaling