Scalable Partial Explainability in Neural Networks via Flexible Activation Functions (Student Abstract)

Schyler Chengyao Sun, Chen Li, Zhuangkun Wei, Antonios Tsourdos, Weisi Guo

AAAI 2021 pp. 15899-15900

doi:10.1609/AAAI.V35I18.17946 /aaai/2021/sun2021aaai-scalable/

Abstract

Current state-of-the-art neural network explanation methods (e.g. Saliency maps, DeepLIFT, LIME, etc.) focus more on the direct relationship between NN outputs and inputs rather than the NN structure and operations itself, hence there still exists uncertainty over the exact role played by neurons. In this paper, we propose a novel neural network structure with Kolmogorov-Arnold Superposition Theorem based topology and Gaussian Processes based flexible activation function to achieve partial explainability of the neuron inner reasoning. The model feasibility is verified in a case study on binary classification of the banknotes.

PDF AAAI Semantic Scholar

Cite

Text

Sun et al. "Scalable Partial Explainability in Neural Networks via Flexible Activation Functions (Student Abstract)." AAAI Conference on Artificial Intelligence, 2021. doi:10.1609/AAAI.V35I18.17946

Markdown

[Sun et al. "Scalable Partial Explainability in Neural Networks via Flexible Activation Functions (Student Abstract)." AAAI Conference on Artificial Intelligence, 2021.](https://mlanthology.org/aaai/2021/sun2021aaai-scalable/) doi:10.1609/AAAI.V35I18.17946

BibTeX

@inproceedings{sun2021aaai-scalable,
  title     = {{Scalable Partial Explainability in Neural Networks via Flexible Activation Functions (Student Abstract)}},
  author    = {Sun, Schyler Chengyao and Li, Chen and Wei, Zhuangkun and Tsourdos, Antonios and Guo, Weisi},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2021},
  pages     = {15899-15900},
  doi       = {10.1609/AAAI.V35I18.17946},
  url       = {https://mlanthology.org/aaai/2021/sun2021aaai-scalable/}
}