This is a fast Singular Spectrum Transformation (SST) implementation. SST can be used for anomaly detection over temporal sequence data.

# g++ -O2 sst.cpp -llapack -lblas
# ./a.out

NOTE: LAPACK and BLAS library is required for compiling.

Please, see sst_test() in sst.cpp and CpuSstWorkspace class in felixsst.h. FelixSstWorkspace implements a fast version SST. NaiveSstWorkspace implements a SVD-based SST.

MIT License.

• データストリーム処理における GPU タスク並列を用いたスケーラブルな異常検知
• Change-Point Detection using Krylov Subspace Learning