Adaptive Hierarchical Hyper-gradient Descent

RENLONG JIE, Junbin Gao, Andrey Vasnev, Minh-Ngoc Tran

Keywords: learning rate adaptation, hyper-gradient descent, meta learning, optimisation, hierarchical system

Abstract:

Adaptive learning rates can lead to faster convergence and better final performance
for deep learning models. There are several widely known human-designed adap-
tive optimizers such as Adam and RMSProp, gradient based adaptive methods
such as hyper-descent and L4, and meta learning approaches including learning
to learn. However, the issue of balancing adaptiveness and over-parameterization
is still a topic to be addressed. In this study, we investigate different levels of
learning rate adaptation based on the framework of hyper-gradient descent, and
further propose a method that adaptively learns the model parameters for combin-
ing different levels of adaptations. Meanwhile, we show the relationship between
adding regularization on over-parameterized learning rates and building combi-
nations of different levels of adaptive learning rates. The experiments on several
network architectures including feed-forward networks, LeNet-5 and ResNet-18/34
show that the proposed multi-level adaptive approach can outperform baseline
adaptive methods in a variety circumstances with statistical significance.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

One-sentence Summary: We introduce hierarchical learning rate adaptation with hyper-gradient descent for deep neural networks.

Reviewed Version (pdf): /references/pdf?id=GFquR4vr43