Keras and modern convnets, on TPUs

GDG Belfast

Join the UK&I DevParty for a workshop where you will learn how to build, train and tune your own convolutional neural networks from scratch with Keras and Tensorflow 2. We will look at multiple approaches from very simple transfer learning to modern convolutional architectures such as Squeezenet, including theoretical explanations about neural networks.

Apr 14, 2021, 5:00 – 8:00 PM (UTC)

2 RSVP'd

Key Themes

Machine Learning

About this event

What you'll learn

- To use Keras and Tensor Processing Units (TPUs) to build your custom models faster.

- To use the tf.data.Dataset API and the TFRecord format to load training data efficiently.

- To use transfer learning instead of building your own models.

- To use Keras sequential and functional model styles.

- To build your own Keras classifier with a softmax layer and cross-entropy loss.

- To fine-tune your model with a good choice of convolutional layers.

- To explore modern convnet architecture ideas like modules, global average pooling, etc.

- To build a simple modern convnet using the Squeezenet architecture.

Organizers

  • Rachel Connor

    GDG Organizer

  • Simon Hewitt

    GDG Organizer

  • Yuan Zhang

    Final Year CS Student

  • imy Greene

Contact Us