imprasshant / MTL

Multi-task Learning

This repository contain code to solve the classification task by learning multiple related task in a shared space to capture shared features among all the tasks. Each task is also assigned a private neural network to capture private features. The Convolution Neural network is used as shared neural network and private neural network.

Files

1. MTL.py: Classification model
2. Task_2_class.csv: Use this when 2 tasks are trained in shared space
3. Task_3_class.csv: Use this when 3 tasks are trained in shared space
4. Task_4_class.csv: Use this when 4 tasks are trained in shared space
5. Task_5_class.csv: Use this when 5 tasks are trained in shared space

Dependencies

Python:2.7.5 Keras:2.2.4 numpy: 1.16.0 gensim: 3.7.0 tensorflow: 1.14.0 nltk:3.4 pandas: 0.23.4