This is code for the UCI experiment in paper "Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks"
Python3, numpy, sklearn
Download and decompress the pre-processed datasets used in paper "Do we need hundreds of classifiers to solve real world classification problems?" by running
bash setup.sh
python UCI.py -max_tot N -max_dep dep -file output_file
Use option -max_tot N to skip datasets with size larger than N.
Use option -max_dep dep to set the maximum depth allowed for NTK.
Use option -file output_file to set the output file.
Compare with other classifiers using results reported by "Do we need hundreds of classifiers to solve real world classification problems?" from the link blow:
Details are listed in paper "Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks".