Part 2 for A disciplined approach to neural network hyper-parameters: Part 1

Deep Learning
1

Hi,
Is there a part2 paper for the A disciplined approach to neural network hyper-parameters: Part 1 – learning rate, batch size, momentum, and weight decay paper?. The paper did indicate there maybe a part 2 to it, but I cant find it on google.
Thanks and regards.

2

There isn’t any as of April 2019.

3

I emailed the author, got:

Dear Eli,

Thank you for your interest in my research.

When I wrote that report I was planning a second part on architectures but a number of papers that appeared that seemed to cover the topic, so there isn’t a part 2.

Two papers I would recommend are:

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks at https://arxiv.org/abs/1905.11946

Designing Network Design Spaces at https://arxiv.org/abs/2003.13678

5 Likes