This package implements the most common randomized matrix computation algorithms.
Download the "Year Prediction Million Song Dataset"
- go to the directory "data/"
- Linux: bash LinuxDownloadData.sh
- Mac: bash MacDownloadData.sh
Convert the data to NumPy data file
- python processLibSVMData.py
Wait a while. The output file is "YearPredictionMSD.npy"
Here are some examples.
-
Matrix sketching
- "sketch/demo/demo_rft.py": matrix coherence after the randomized Fourier transform (RFT) gets much smaller.
- "sketch/demo/demo_sketch.py": apply SRFT, count sketch, and leverage score sampling to matrix multiplication and compare their errors.
-
Optimization
- "optimization/demo/demo_precondition_cg": the converge of CG with/without preconditioning.