• quora_q_pairs_v0_1.py: kaggle上的一个kernal的原始代码

• quora_q_pairs_v0_2.py: 使用dot代替concatenate + 多层dense

• quora_q_pairs_v0_3.py: 试试bidirectional lstm + dropout

• quora_q_pairs_v0_4.py: deep bidirectional lstm + no dropout

• quora_q_pairs_v0_5.py: deep bidirectional lstm(shared both lstm and bidirection) + dropout + concatenate

• quora_q_pairs_v0_6.py: Pool 预处理text

• quora_q_pairs_v0_7.py: name_scope + GRU + attention

• quora_q_pairs_v0_8.py: bidirectional GRU + attention + stack

• predict_quora_q_pairs_v0_1.py: 没有model.fit只是预测

• test_build_model_v0_1.py: 用随机数测试模型 - 3层bilstm + concate + 2fc

• test_build_model_v0_2.py: 用Sequential自定义的layer的tensorboard test

• test_build_model_v0_3.py: tensorboard测试, 清晰定义name_scope

• XGB_handcrafted_leaky.py: kaggle上0.158得分的XGB程序

数据

输出