NLP Status Report 2017-6-5 - 版本历史

2017年6月5日 (一) 06:04 Zhangsy

2017-06-05T06:04:02Z

2017年6月5日 (一) 06:01 Zhangsy

2017-06-05T06:01:33Z

2017年6月5日 (一) 05:56 Zhangsy

2017-06-05T05:56:30Z

2017年6月5日 (一) 05:52 Zhangsy

2017-06-05T05:52:56Z

2017年6月5日 (一) 05:45 Intern

2017-06-05T05:45:52Z

2017年6月5日 (一) 05:45 Intern

2017-06-05T05:45:08Z

2017年6月5日 (一) 05:44 Intern

2017-06-05T05:44:46Z

2017年6月5日 (一) 05:35 Intern

2017-06-05T05:35:37Z

Intern：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/5/31 |Jiyuan Zhang || || |- |Aodong LI || * Small data: Only make the English...”为内容创建页面

2017-06-05T05:34:54Z

以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/5/31 |Jiyuan Zhang || || |- |Aodong LI || * Small data: Only make the English...”为内容创建页面

新页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/5/31
|Jiyuan Zhang ||
||
|-
|Aodong LI ||
* Small data:
Only make the English encoder's embedding constant -- 45.98
Only initialize the English encoder's embedding and then finetune it -- 46.06
Share the attention mechanism and then directly add them -- 46.20
* big data baseline bleu = '''30.83'''
* Fixed three embeddings
{| class="wikitable"
|-
! alpha
! beta
! result (bleu)
|-
| 1
| 1
| 43.50
|-
| 4/3
| 2/3
| 43.58 (w/o retrained)
|-
| 2/3
| 4/3
| 41.22 (w/o retrained)
|-
| 2/3
| 4/3
| 42.36 (w/ retrained)
|}
* experiments with '''constant''' initialized embedding:

|}

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||

||

|-

|}

@@ 第42行： / 第42行： @@
 * reviewing related papers
 ||
+* solve EOS problem
 |-
 |Shipan Ren ||

@@ 第37行： / 第37行： @@
 | 36.58, 58.4/40.4/32.1/27.0 BP=0.968
 |}
 ||

@@ 第23行： / 第23行： @@
 * trained word2vec on big data, and directly used it on NMT, but resulted in quite poor performance
 * trained M-NMT model, got bleu=36.58 (+1.34 than NMT). But found the EOS in mem has a big influence on result:
+{| class="wikitable"
 ||

@@ 第12行： / 第12行： @@
    Share the attention mechanism and then directly add them -- 46.20
 * big data baseline bleu = '''30.83'''
-* Fixed three embeddings
+* Model with three fixed embeddings
    Shrink output vocab from 30000 to 20000 and best result is 31.53
    Train the model with 40 batch size and best result until now is 30.63

@@ 第15行： / 第15行： @@
    Shrink output vocab from 30000 to 20000 and best result is 31.53
    Train the model with 40 batch size and best result until now is 30.63
-|}
 * test more checkpoints on model trained with batch = 40
 * train model with reverse output
 |-
 |Shiyue Zhang ||

@@ 第2行： / 第2行： @@
 !Date !! People !! Last Week !! This Week
 |-
-| rowspan="6"|2017/5/31
+| rowspan="6"|2017/6/5
 |Jiyuan Zhang ||
 ||
@@ 第13行： / 第13行： @@
 * big data baseline bleu = '''30.83'''
 * Fixed three embeddings
-{| class="wikitable"
+  Shrink output vocab from 30000 to 20000 and best result is 31.53
-|-
+  Train the model with 40 batch size and best result until now is 30.63
-! alpha
-! beta
-! result (bleu)
-|-
-| 1
-| 1
-| 43.50
-|-
-| 4/3
-| 2/3
-| 43.58 (w/o retrained)
-|-
-| 2/3
-| 4/3
-| 41.22 (w/o retrained)
-|-
-| 2/3
-| 4/3
-| 42.36 (w/ retrained)
 |}
-* experiments with '''constant''' initialized embedding:
+* test more checkpoints on model trained with batch = 40
+* train model with reverse output
 ||