AI Saturdays | SGInnovate
January 13
2018

Location

32 CARPENTER STREET, Singapore 059911
Singapore

AI Saturdays

Presented by SGInnovate

Join AI researchers, engineers, data scientists and self-learners going in-depth into materials like actual recorded lecture videos and research paper readings from top universities like Stanford and UC Berkeley – covering practical deep learning, in-depth deep learning theory from multiple perspectives, Reinforcement Learning, computer vision and natural language processing. Free to attend!

Nurture.AI is partnering with SGInnovate to organise AI Saturdays in Singapore, which will be held at its premises at 32 Carpenter Street, and BASH at 79 Ayer Rajah Crescent (starting from March 2018). 

 

In order to cater to a diverse audience, there will be 3 structured sessions every Saturday – you can attend all, some or none, it’s totally up to you! If you don’t want to attend some of the sessions, throughout the day there will be open hacking on creating open-source code implementations of the top research paper pre-prints that week. Or use that time to catch-up on lectures and readings (sessions 2 and 3 have many hardcore readings by the way!) while discussing with peers.

Session 1: 10am - 12pm – Practical Deep Learning (Beginner-Intermediate)
12-1pm – Lunch (occasional brown bag lunch talk from an expert)
Session 2: 1pm - 3pm – Deep Learning Theory (Intermediate-Advanced)
Session 3a: 3pm - 6pm – Reinforcement Learning (Intermediate-Advanced)
Session 3b: 3pm - 6pm – Convolutional Neural Networks for Visual Recognition (Intermediate-Advanced)
Session 3c: 3pm - 6pm – Natural Language Processing With Deep Learning (Intermediate-Advanced)

Community Rules and Philosophy
The course materials we choose are widely known to be clear in explanations, and the presenters are leading expert authorities in the field, so we watch the original lectures directly instead of recreating our own in the meetup sessions
The value of coming together as a study group lies in clarifying doubts, deeper discussions into the material, and accountability on course completion – we are clear about that in the way we structure the activities that happen in the study group. This is a judgement-free, safe zone, and people who have promised to prepare are expected to do so in order to ensure the session is fruitful for everyone.
We constantly co-create and iterate on improving the learning experience together! Because all these sessions are free, we rely heavily on the community’s participation and support to make this work.

What do we do in the Sessions
Session 1: Practical Deep Learning 
Using the free and proven Fast.ai materials, this is perfect for beginners in deep learning and machine learning, with some prior Python programming experience and high school math knowledge – and it’d get you to a stage where you can implement cutting-edge deep learning models, in just 14 weeks! No worries if you have no Python programming experience, feel free to reach out and we’d be happy to advise on what you can use in the weeks leading up to the start date to prepare – you can certainly get up to speed if you work hard in these few weeks, but time is running short so get started now!

We will watch the lectures as a group, stop the video for discussion at any point if anyone has a question, and also breakout into small groups for the in-lecture exercises – removing any obstacles along the way, making sure that you can progress through the course confidently if you stick with us – that’s our commitment to you for your time investment!

Session 2: Deep Learning Theory
We start off with materials from the Stanford STAT385 course on Theories of Deep Learning. For this particular session, the true value of the physical meetup lies in discussing the theoretically-dense research paper readings. 

Participants are expected to have viewed the lecture video for the week beforehand, and each participant will take charge of being the expert authority on one of the readings for the week in the discussion by having thoroughly read and researched it, to make the best use of everyone’s time. You can help each other prepare better by posting your questions on specific parts of the papers using the Nurture.AI platform’s highlight-commenting function.

Session 3a: Reinforcement Learning
We start off with materials from David Silver’s UCL/DeepMind Reinforcement Learning course, before continuing with UC Berkeley CS294 Deep Reinforcement Learning. We will view the lecture as a group, stop the video for discussion at any point if anyone has a question, and breakout into small groups to discuss papers mentioned. We will then end off with code practice on implementing the techniques covered in the session, which can then be completed over the week. 

Session 3b: Convolutional Neural Networks for Visual Recognition
Covering the material in Stanford’s CS231n Spring 2017 course headed by Prof Fei-Fei Li (Chief AI Scientist of Google Cloud, Director of Stanford AI Lab), we will take the first 1.5 hours to view and discuss the lecture together. We will take another half an hour to discuss the specifics of the paper readings for the week and clarify questions. The last hour will be dedicated to kicking off the participant’s implementation of the models discussed, which can be completed over the following week. This session will cover topics like image classification, object detection, image caption, visual question answering, feature visualisation and adversarial training.

For ease of the facilitator in-charge’s preparation for any particular week, participants are encouraged to read the papers beforehand and post questions on specific parts of the papers using the Nurture.ai platform’s highlight-commenting function.

Session 3c: Natural Language Processing with Deep Learning
Covering the material in Stanford’s CS224n Winter 2017 course taught by Professor Christopher Manning and Richard Socher (Chief Scientist of Salesforce), we will take the first 1.5 hours to view and discuss the lecture together. We will take another half an hour to discuss the specifics of the paper readings for the week and clarify questions. The last hour will be dedicated to kicking off the participant’s implementation of the models discussed, which can be completed over the following week. This session will cover topics like word vector representations, dependency parsing, recurrent neural networks and language models, machine translation, attention models, tree recursive neural networks, and speech processing.

For ease of the facilitator in-charge’s preparation for any particular week, participants are encouraged to read the papers beforehand and post questions on specific parts of the papers using the Nurture.ai platform’s highlight-commenting function.

 

Program for 13/01/18
Session 1: Fast.ai Lesson 3 – Why Deep Learning; Intro to Convolutions
Session 2: Stat385 Lecture 3 Readings – Harmonic Analysis of CNNs
Session 3a: UCL/Deep Mind Reinforcement Learning Lecture 2 – Markov Decision Processes
Session 3b: Loss Functions and Optimization
Session 3c: Advanced Word Vector Representations

 

Topics: Artificial Intelligence / Deep Learning / Machine Learning / Robotics

You may also like the following: