Dynamic Fusion of LSTM Predictions Using Reinforcement Learning-Based Gowla for Human Activity Recognition | Synapse