Federated Learning for Sentiment Analysis using Neural Networks
Published:
As a part of Minor Project we tried to integrate two different research fields namely Federated Learning and Sentiment Analysis. We experimented with a newly open sourced (at the time of our project) tensorflow-federated library to see the impact of federated training setup on sentiment analysis task. Special thanks to my teammates Vinay, Karthik and mentor Dr. Manu Vardhan for their support.
Data: We extracted the data from sentiment140 and modified it according to the federated data format.
Algorithm: BiLSTM, ELMO, Federated averaging
Major tool(s): Tensorflow, Keras, Tensorflow-federated, NLTK, Numpy, Pandas
An image of the web UI of the project:
Loss VS Rounds (Different than epochs):
