AIMarkerFinder is a computational tool designed for the analysis of high-dimensional biomedical data (e.g., metabolomic or genomic datasets) to identify minimal sets of informative biomarkers. The method combines a denoising autoencoder with an attention mechanism (DAE) and a Kolmogorov–Arnold Network (KAN) to perform feature selection and build interpretable classification models. This approach effectively addresses the "curse of dimensionality" and feature redundancy, enhancing both model accuracy and interpretability.
- Python ≥ 3.10
- Poetry for dependency management
Install dependencies using Poetry:
poetry installView all available arguments:
poetry run aimf -hBasic command syntax:
poetry run aimf <input_file> [options]Supported input formats: .csv, .tsv, .xls, .xlsx
Class labels: Specify with -CC; if omitted, all non-numeric columns are used by default.
Example usage with an input file example/test_data.xlsx:
poetry run aimf example/test_data.tsv -balance -CC Diagnose