LossPlot: A Better Way to Visualize Loss Landscapes

Robert Bain; Mikhail Tokarev; Harsh Kothari; Rahul Damineni

LossPlot: A Better Way to Visualize Loss Landscapes

Machine Learning 2021-12-01 v1 Computer Vision and Pattern Recognition Human-Computer Interaction

Authors: Robert Bain , Mikhail Tokarev , Harsh Kothari , Rahul Damineni

Abstract

Investigations into the loss landscapes of deep neural networks are often laborious. This work documents our user-driven approach to create a platform for semi-automating this process. LossPlot accepts data in the form of a csv, and allows multiple trained minimizers of the loss function to be manipulated in sync. Other features include a simple yet intuitive checkbox UI, summary statistics, and the ability to control clipping which other methods do not offer.

Keywords

image processing image classification visualization

Cite

@article{arxiv.2111.15133,
  title  = {LossPlot: A Better Way to Visualize Loss Landscapes},
  author = {Robert Bain and Mikhail Tokarev and Harsh Kothari and Rahul Damineni},
  journal= {arXiv preprint arXiv:2111.15133},
  year   = {2021}
}

Comments

5 pages; 2 large figures

Related papers

View all related →