2019"默克"杯逆合成反应预测大赛
Model: Transformer/RNN
Framework: tensor2tensor/opennmt-py
Data format: SMILES
Visualize Tool: RDKit
Evaluation Method is the same as SQuAD dataset, code here
| modelName | framework | score | f1 | em |
|---|---|---|---|---|
| transformer-base | tensor2tensor | 0.627 | 0.764 | 0.218 |
| transformer-merge | tensor2tensor | 0.636 | 0.770 | 0.235 |
| transformer-base | opennmt-py | 0.646 | 0.860 | 0.002 |
| transformer-base-200000 | opennmt-py | 0.650 | 0.865 | 0.004 |
| transformer-base-254000 | opennmt-py | 0.653 | 0.869 | 0.005 |
| transformer-base-400000 | opennmt-py | 0.660 | 0.877 | 0.007 |
| transformer-base-800000 | opennmt-py | 0.664 | 0.881 | 0.013 |
| rnn-based-115000 | opennmt-py | 0.415 | 0.554 | 0.000 |
| modelName | framework | metric | score | f1 | em |
|---|---|---|---|---|---|
| rnn-based-6000 | opennmt-py | char-based | 0.500 | 0.663 | 0.000 |
| rnn-based-48000 | opennmt-py | char-based | 0.505 | 0.673 | 0.0015 |
| rnn-based-48000 | opennmt-py | reactants-based | 0.0098 | 0.0126 | 0.0015 |
| rnn-based-big-24000 | opennmt-py | reactants-based | 0.0139 | 0.0185 | 0.0000 |
1.《Retrosynthetic Reaction Prediction Using Neural Sequence-to-Sequence Models》
2.《Found in Translation Predicting Outcomes of Complex Organic Chemistry Reactions using Neural Sequence to Sequence Models》
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent