Keras: An Easy Way to Build Deep Learning Model
- 2017-04-25 (Tue.), 14:00 PM
- Recreation Hall, 2F, Institute of Statistical Science
- Dr. Cheng-Han Tu
- Department of Mathematics, National Taiwan University
Abstract
Fast prototyping and modification of a neural network are essential to competitive deep learning design and model deployment. Among many deep learning frameworks, Keras is a user-friendly tool for model design and modification. ? Based on python script and Theano/TensorFlow as backends, Keras also seamlessly support GPU acceleration. In this lecture, we will provide several examples to visualize deep learning model improvement. ? Keras scripts from choosing optimizer to overfitting prevention are introduced. Finally, a basic convolution neural network with GPU acceleration will be demonstrated.
Update:2024-12-10 17:06