5-way Few-shot Classification on Penn Treebank v1 (test)
72.81-shot Accuracy (random)LSTM-LM oracle
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LSTM-LM oracle2017.12 | 72.8 | 72.8 | 72.8 | 72.8 | 72.8 | 72.8 | |
| adaLSTMConditioning information=error gradient (V), Number of layers=22017.12 | 43.1 | 43 | 52.05 | 54.2 | 57.35 | 58.4 | |
| adaLSTMConditioning information=direct feedback (DF), Number of layers=22017.12 | 41.25 | 43.2 | 52.1 | 52.9 | 57.8 | 58.8 | |
| adaLSTMConditioning information=error gradient (V), Number of layers=12017.12 | 36.55 | 37.7 | 43.25 | 44.6 | 50.7 | 52.1 | |
| adaLSTMConditioning information=direct feedback (DF), Number of layers=12017.12 | 36.35 | 36.3 | 41.6 | 43.4 | 49.1 | 50.1 | |
| LSTM + adaFFNConditioning information=direct feedback (DF), Number of layers=22017.12 | 33.65 | 35.3 | 46.6 | 47.8 | 51.4 | 52.6 | |
| LSTM + adaFFNConditioning information=error gradient (V), Number of layers=22017.12 | 32.55 | 33.2 | 44.15 | 46 | 50.4 | 51.7 | |
| Matching Nets2017.12 | 32.4 | — | 36.1 | — | 38.2 | — |