Log In Sign Up

Which Minimizer Does My Neural Network Converge To?

by   Manuel Nonnenmacher, et al.

The loss surface of an overparameterized neural network (NN) possesses many global minima of zero training error. We explain how common variants of the standard NN training procedure change the minimizer obtained. First, we make explicit how the size of the initialization of a strongly overparameterized NN affects the minimizer and can deteriorate its final test performance. We propose a strategy to limit this effect. Then, we demonstrate that for adaptive optimization such as AdaGrad, the obtained minimizer generally differs from the gradient descent (GD) minimizer. This adaptive minimizer is changed further by stochastic mini-batch training, even though in the non-adaptive case GD and stochastic GD result in essentially the same minimizer. Lastly, we explain that these effects remain relevant for less overparameterized NNs. While overparameterization has its benefits, our work highlights that it induces sources of error absent from underparameterized models, some of which can be challenging to control.


page 1

page 2

page 3

page 4


Weighted Neural Tangent Kernel: A Generalized and Improved Network-Induced Kernel

The Neural Tangent Kernel (NTK) has recently attracted intense study, as...

Input Invex Neural Network

In this paper, we present a novel method to constrain invexity on Neural...

Nondeterminism and Instability in Neural Network Optimization

Nondeterminism in neural network optimization produces uncertainty in pe...

Some like it tough: Improving model generalization via progressively increasing the training difficulty

In this work, we propose to progressively increase the training difficul...

Do We Need Zero Training Loss After Achieving Zero Training Error?

Overparameterized deep networks have the capacity to memorize training d...