“Simplifying Neural Networks using Formal Verification” by Sumathi Gokulanathan, Alexander Feldsher, Adi Malca, Clark Barrett, and Guy Katz. In NASA Formal Methods: 12th International Symposium, (NFM '20), (Ritchie Lee, Susmit Jha, and Anastasia Mavridou, eds.), May 2020, pp. 85-93. Moffet Field, California.
Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10%.
BibTeX entry:
@inproceedings{GFM+20, author = {Sumathi Gokulanathan and Alexander Feldsher and Adi Malca and Clark Barrett and Guy Katz}, editor = {Ritchie Lee and Susmit Jha and Anastasia Mavridou}, title = {Simplifying Neural Networks using Formal Verification}, booktitle = {{NASA} Formal Methods: 12th International Symposium, (NFM '20)}, series = {Lecture Notes in Computer Science}, pages = {85--93}, publisher = {Springer}, month = may, year = {2020}, isbn = {978-3-030-55754-6}, note = {Moffet Field, California}, url = {http://theory.stanford.edu/~barrett/pubs/GFM+20.pdf} }
(This webpage was created with bibtex2web.)