Table 5:
Clustering performance. CAT-LSTM denotes our CAT-LSTM model jointly trained with auto-encoder. Note that higher CHI and lower DBI relate to a model with better separation between the clusters.
CHI ↑ | DBI ↓ | |
---|---|---|
CNN+LSTM | 1.40(±0.32) | 0.93(±0.12) |
CAT-LSTM-v1 | 1.82(±0.24) | 0.75(±0.09) |
CAT-LSTM-v2 | 1.75(±0.26) | 0.88(±0.10) |