This is a package for the implementation of the Multi-Layer Kernel Machine (MLKM), which is a framework for multi-scale nonparametric regression and confidence bands. The method integrates random feature projections with a multi-layer structure.
pip install Multi-Layer-Kernel-Machine
See MLKM package in PyPi.
- Python 3
- Pytorch
- numpy, pandas, matplotlib, tqdm, scikit-learn
See MLKM documentation.
Multi-Layer Kernel Machine (MLKM) is released under the MIT License.
Datasets | Instances | Attributes | Source |
---|---|---|---|
YearPredictionMSD | 515,345 | 90 | Download through YearPredictionMSD |
SML2010 | 4,137 | 20 | Download through SML2010 |
DryBean | 13,611 | 16 | Download through Dry Bean Dataset |
Run Simulation Examples directly.
To obtain the results in Example 1, run Sparse Additive - d=x.ipynb
for x
To obtain the results in Example 2, run ATLAS model - d=x.ipynb
for x
To obtain the results in Example 3, run Additive (function) (distribution).ipynb
for additive models with function
distribution
To obtain the results in Example 4, run Example4.ipynb
.
To obtain the results in Appendix, run codes in Additional Examples
.
Run Real Data Examples after downloading and unzipping the data. Store the data and codes in the same path and don't change the filenames of them.
To obtain the results of temperature forecasting data, run RealData_SML.ipynb
.
To obtain the results of audio feature data, run RealData_MSD.ipynb
for small-scale data and run RealData_large_MSD.ipynb
for large-scale data.
To obtain the results in Appendix, run RealData_DryBean.ipynb
.
Our experiments can also be carried out using our Multi-Layer-Kernel-Machine package. We provide an example for dataset MSD as a quick start.