The Deep Learning Workshop organized by the Hong Kong Society of Artificial Intelligence and Robotics (HK-SAIR) and co-organized by HKUST Big Data Institute (BDI). The workshop will be held on 30 April 2019 (Tue) from 2:00-5:00pm at Cheung On Tak Lecture Theatre (LT-E) in HKUST.
In this session, we will dive into the history about the evolution of different CNN network architectures from classical LeNet to the-state-of-art SENet.
First of all, we will show the connection between the Support Vector Machine and Deep Neural Network. Then we will give an introduction of CNN and interpret it as the regularized form of MLP. Furthermore, we will review, compare and analyze different CNN milestones by following the timeline. The audiences will gain a deeper understanding of the CNN model design and see what kind of differences has been made.
You will learn how to build convolutional neural networks from scratch with Tensorflow2.0 and apply it to classify number images. TensorFlow 2.0 is a significant milestone and a major new release focusing on ease of use and simplification. With Keras tightly integrated and eager execution by default, it aligns users’ expectations about the programming model better with TensorFlow practice and should make TensorFlow easier to learn and apply. During the workshop, you will get started with Tensorflow2.0 Alpha, an early version released, for now, to try the new way to build an image classification model with several convolutional neural layers.
Date: 30 April 2019 (Tuesday)
Venue: Cheung On Tak Lecture Theatre (LT-E), HKUST
2:00pm Welcome Speech
1. Extend from Logistic Regression
Depth: Deep NIN 2. CNN (CV)
(1) From MLP to LeNet (Spatial Locality) (2) AlexNet (From CPU2GPU)
(3) VGG (3*3 conv)
(4) Inception (Multi-scale)
(5) Resnet, Densnet (Highway)
(6) SE-NET (Channel Independence)
3:00pm Coffee Break
3:30pm Hands-on Lab Tutorial*
5:00pm Q & A
Quota: 100 people per workshop
* Participants are required to bring their own laptop for the lab session
This is an open-to-all event on a first-come-first-served basis. To reserve a place, register now at here. More workshop details are also available on the registration site.
HK-SAIR is a non-profit organization focuses on Artificial Intelligence and Robotics (AIR) research and the AIR industry development. It cooperates with top universities and research committees in Hong Kong, making contributions to the innovations in Hong Kong’s financial industry, trade logistics, industrial manufacturing, supply chain and other industries.
If you have any enquiries, please feel free to contact [email protected] / [email protected].