Recurrent Neural Network-Based Architecture

The performance of existing CNN based architectures is poor for small models as they distort spatial resolution. Due to this drawback, they are difficult to implement for real-time sequence-learning problems. Backpropagation through time (BPTT) algorithm is implemented in recurrent neural network for learning. The information present in the output node is dependent on the information stored in previous neurons. The classification models are based on long short-term memory (LSTM) which is used for processing and making predictions.

Akshay Juneja authored 15+ articles for INFO4EEE Website on Deep Learning.

No comments

Powered by Blogger.