TCN-with-attention-master
所属分类：其他
开发工具：Python
文件大小：16KB
下载次数：26
上传日期：2020-06-03 14:51:33
上 传 者：
对我而言可
说明： 基于注意力的的方法，预测使用的tcn，tcn是比lstm更好的一种预测方法，附数据
(Based on the attention method, TCN is better than LSTM in prediction, with data attached)
文件列表:
data (0, 2018-05-31)
dataset.py (1926, 2018-05-31)
kor_char_parser.py (2776, 2018-05-31)
main.py (6164, 2018-05-31)
results (0, 2018-05-31)
results\without_attention.txt (21556, 2018-05-31)
results\with_attention.txt (21586, 2018-05-31)
tcn.py (4598, 2018-05-31)
# TCN with attention
Temporal Convolutional Network with attention layer
Concept of model is mostly like [Simple Neural Attentive Meta-Learner (SNAIL)](https://github.com/sagelywizard/snail).
But in this model, attention layer is on every top of convolutions layers. And attention size is differ from SNAIL.
## Result
Dataset: [Agnews](https://github.com/mhjabreel/CharCNN/tree/master/data/ag_news_csv) without pre-processing
- with attention: 0.82
- without attention: 0.81
### My thoughts on result
Most of simple models on agnews shows 0.81 accuracy. (Which tested on [A Structured Self-Attentive Sentence Embedding](https://github.com/flrngel/Self-Attentive-tensorflow), [TagSpace](https://github.com/flrngel/TagSpace-tensorflow) and it uses word based embedding)
So 0.82 accuracy with **character based model** seems worthiness.
近期下载者：
相关文件：
收藏者：