(*) Insect cyborgs: Bio-mimetic feature generators improve machine learning accuracy on limited data. NeurIPS 2019
This paper describes experiments in which a computational model of a moth brain (MothNet) acts as fully-automated, front-end feature generators for ML methods. The MothNet-generated features substantially improve ML accuracy on vectorized MNIST and Omniglot datasets. The MothNet features also out-perform comparison feature generators such as PCA, PLS, and NNs.
The poster presented at a NeurIPS 2019 workshop is here neurips poster.
The full paper is available on arXiv at insect cyborgs.
A local copy is here.
