chaosmining-dataset

Chaosmining is a synthetic dataset designed to evaluate post-hoc local attribution methods in low signal-to-noise ratio environments. It is created and maintained by Davidson Lab at UC Davis. The code to generate data and run experiments can be found at "https://github.com/geshijoker/ChaosMining". The dataset consists of "Symbolic Functional Data", "Vision Data", and "Audio Data".

The symbolic functional data is created based on human-designed symbolic functions with ground truth annotations derived from mathematical formulas. It is used to study the general behaviors of multilayer perceptron (MLP) networks on regression tasks. 

The vision data is used to study popular architectures for visual scene classification tasks in various noise conditions. 

The audio data is used to study popular architectures for sequential data classification tasks in various noise conditions. 

These synthetic datasets serve as accessible resources for researchers exploring feature selection and attribution methods in the context of explainable artificial intelligence (XAI).