“Work-progress 2017.08”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
（以“===Daily Report=== {|class="wikitable" ! Date !! Person !! start!! leave !! hours !! plan !!status |- | rowspan="1"|2017.8.3 |Xin Jing ||10:00A.M|| 7P.M|| 8h ||st...”为内容创建页面）
 
（→‎Daily Report）
第4行： 第4行：
 
! Date !! Person  !! start!! leave !! hours !! plan !!status
 
! Date !! Person  !! start!! leave !! hours !! plan !!status
 
|-
 
|-
−
| rowspan="1"|2017.8.3
+
| rowspan="1"|2017.8.1
−
|Xin Jing ||10:00A.M|| 7P.M||  8h ||study RNN|| look for possible ways to implement our model
+
|Xin Jing ||10:00A.M|| 7P.M||  8h || change the batch size of the most basic model to 1 (the model with only past returns as inputs) ||Amazingly, the results became very stable, and with minor adjustment of number of hidden states, our RNN model outperforms base line model markedly.
 
|-
 
|-
−
| rowspan="1"|2017.7.4
+
| rowspan="1"|2017.8.2
−
|Xin Jing ||10:00A.M|| 6P.M||  8h ||study RNN|| look for possible ways to implement our model
+
|Xin Jing ||10:00A.M|| 7P.M||  8h || add DCC-bivariate Garch as a baseline || implementing DCC-bivariate GARCH
 
|-
 
|-
−
| rowspan="1"|2017.7.7
+
| rowspan="1"|2017.8.3
−
|Xin Jing ||10:00A.M|| 9P.M||  11h ||implement RNN for our model|| implement an RNN with inputs being past returns
+
|Xin Jing ||10:00A.M|| 7P.M||  8h || add DCC-bivariate Garch as a baseline || implemented DCC-bivariate GARCH
−
|-
+
−
| rowspan="1"|2017.7.10
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||implement an RNN with inputs being past returns|| implementing
+
−
|-
+
−
| rowspan="1"|2017.7.13
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||implement an RNN with inputs being past returns|| implementing
+
−
|-
+
−
| rowspan="1"|2017.7.14
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||implement an RNN with inputs being past returns|| implementing
+
−
|-
+
−
| rowspan="1"|2017.7.17
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||realized that base line in python does not converge || re-implemented baseline in R
+
−
|-
+
−
| rowspan="1"|2017.7.18
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||realized that base line in python does not converge || re-implemented baseline in R
+
−
|-
+
−
| rowspan="1"|2017.7.21
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||realized that base line in python does not converge || re-implemented baseline in R
+
−
|-
+
−
| rowspan="1"|2017.7.24
+
−
|Xin Jing ||10:00A.M|| 7P.M||  9h || comparing results from RNN with those from baseline || very close results but results from RNN not stable
+
 
|-
 
|-
−
| rowspan="1"|2017.7.25
+
| rowspan="1"|2017.8.7
−
|Xin Jing ||10:00A.M|| 7P.M||  9h || changing parameters(hidden layer, time steps) and compare results || still, results good but not stable
+
|Xin Jing ||10:00A.M|| 7P.M||  8h ||proof reading for chapter 1 of Professor Wang's book || proof reading for chapter 1 of Professor Wang's book
 
|-
 
|-
−
| rowspan="1"|2017.7.26
+
| rowspan="1"|2017.8.9
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||changing features|| try features with past changes of return and past basis. still, results good but not stable
+
|Xin Jing ||10:00A.M|| 7P.M||  8h ||proof reading for chapter 1 of Professor Wang's book || proof reading for chapter 1 of Professor Wang's book
 
|-
 
|-
−
| rowspan="1"|2017.7.27
+
| rowspan="1"|2017.8.10
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||changing features|| try features with absolute values of past returns. still, results good but not stable
+
|Xin Jing ||10:00A.M|| 7P.M||  8h ||correcting chapter 3 of Professor Wang's book || correcting chapter 3 of Professor Wang's book
 
|-
 
|-
−
| rowspan="1"|2017.7.28
+
| rowspan="1"|2017.8.11
−
|Xin Jing ||10:00A.M|| 7P.M||  9h ||changing features|| try features with past returns and moving variances, still, results good but not stable; try features with nothing but all ones, still, results good but not stable
+
|Xin Jing ||10:00A.M|| 7P.M||  8h ||correcting chapter 3 of Professor Wang's book || correcting chapter 3 of Professor Wang's book
 
|-
 
|-
−
 
 
|}
 
|}

2017年8月17日 (四) 04:27的版本

Daily Report

Date Person start leave hours plan status
2017.8.1 Xin Jing 10:00A.M 7P.M 8h change the batch size of the most basic model to 1 (the model with only past returns as inputs) Amazingly, the results became very stable, and with minor adjustment of number of hidden states, our RNN model outperforms base line model markedly.
2017.8.2 Xin Jing 10:00A.M 7P.M 8h add DCC-bivariate Garch as a baseline implementing DCC-bivariate GARCH
2017.8.3 Xin Jing 10:00A.M 7P.M 8h add DCC-bivariate Garch as a baseline implemented DCC-bivariate GARCH
2017.8.7 Xin Jing 10:00A.M 7P.M 8h proof reading for chapter 1 of Professor Wang's book proof reading for chapter 1 of Professor Wang's book
2017.8.9 Xin Jing 10:00A.M 7P.M 8h proof reading for chapter 1 of Professor Wang's book proof reading for chapter 1 of Professor Wang's book
2017.8.10 Xin Jing 10:00A.M 7P.M 8h correcting chapter 3 of Professor Wang's book correcting chapter 3 of Professor Wang's book
2017.8.11 Xin Jing 10:00A.M 7P.M 8h correcting chapter 3 of Professor Wang's book correcting chapter 3 of Professor Wang's book
取自“http://index.cslt.org/mediawiki/index.php?title=Work-progress_2017.08&oldid=28278”