Uninorm-like parametric activation functions for human-understandable neural models
This work addresses interpretability in neural networks for domains requiring transparent decision-making, but it appears incremental as it builds on existing fuzzy logic and MCDM concepts.
The paper tackled the problem of finding human-understandable connections between input features in neural networks by introducing a parameterized activation function based on nilpotent fuzzy logic and multi-criteria decision-making, with the learnable parameter indicating compensation levels, and demonstrated its effectiveness on UCI classification datasets.
We present a deep learning model for finding human-understandable connections between input features. Our approach uses a parameterized, differentiable activation function, based on the theoretical background of nilpotent fuzzy logic and multi-criteria decision-making (MCDM). The learnable parameter has a semantic meaning indicating the level of compensation between input features. The neural network determines the parameters using gradient descent to find human-understandable relationships between input features. We demonstrate the utility and effectiveness of the model by successfully applying it to classification problems from the UCI Machine Learning Repository.