Yang Zhang (Work Log)

From edegan.com
Revision as of 16:58, 28 September 2017 by Yangzhang (talk | contribs)
Jump to navigation Jump to search

09/19/2017 accounts setup and get familiar with the rules

09/21/2017 start to build two different kinds of deep neural networks (Convolutional and LSTM) to classify the companies' industry, specifically for the file "Descriptions.txt"

09/22/2017 the two models both can achieve 90+% training accuracy and ~60% testing accuracy. Notice that the task is hard even for humans and thus 60% is acceptable, plus the baseline is under 10% for random guessing

09/26/2017 modify the previous model and train on the "longdescriptionindu.txt"

09/28/2017 the LSTM model achieves 67% testing accuracy