2015年10月8日 (四) 10:04的版本

Chaos Work

1. Write Matrix Factorization program. 2. Write DNN Max-Margin program with dropout. 3. Run some test model in GPU machines.

<Written by 10-08>

@@ 第1行： / 第1行： @@
 =Chaos Work=
-==Binary Word Vector==
+==Date Line==
-===Reproduce Nested Dropout===
-Nested dropout method proposed by Rippel et. in their paper "Learning Ordered Representations with Nested Dropout", they proposed a dropout method which could learning ordered information in different dimensions.
-====Simple semi-linear autoencoder====
+===Date 2015-10-09===
-Their first draft work is semi-linear autoencoder, so I will reproduce this work.
+====Plan to do====
+. Write Matrix Factorization program.
-And I will compare this work to PCA.
+. Write DNN Max-Margin program with dropout.
+. Run some test model in GPU machines.
-We only consider one hidden layer.
+ <Written by 10-08>
-Start at 2015-07-02 20:00