An AI Evening Lecture

I am pleased to share my introductory lecture on Deep Learning, to provide an overview of the subject, coming from an industrial perspective to 20 PhD and Masters students from Universiti Putra Malaysia.

I would like to thank Dr @Erzam Marlisah for inviting me to the talk and would like to express my delight in finding out that his students, from the questions and arguments that they set forth, had an excellent grasp of the depth and breadth of the subject. Consequently, the lecture, estimated to run for 4 hours, actually ran for about 7 hours.

I hope that others too may benefit from this lecture.

Slides presented during the lecture may be downloaded from the post above.

CONTENTS

Overall, this was an introductory lecture to Deep Learning. However, it was not meant to be just a cursory lecture that only scratches the surface of the subject. In designing what in my opinion an introductory lecture should be, I have tried to provide the foundations that I felt students need to have in order to be able to better grapple with the subject as they proceed further with it. Specifically, that means this lecture attempts to explain the why's along with the normal and customary how's of Artificial Neural Networks, including:

• How the technology domain of AI is usually framed and organized
• How AI works
• What are the major processing steps involved
• Why do we perform those steps, what are the intents behind those steps and mathematical models
• How is AI actually implemented in the real world
• Why is AI important in our current world
• How can AI be used
• How can each student excel and differentiate him/herself from others in the field

PART ONE

In PART 1 of the talk (above), I

• set out the objectives of the lecture 
• covered the general terrain of Artificial Intelligence
• provided a brief description of the important areas of AI
• contrasted the major areas against one another
• introduced popular AI platforms that can be accessed and used for free

PART TWO

In PART 2 of the talk, we covered

• the different characteristics of a few primary types of AIs
• what they are used for
• how they work in broad and general terms
• then we moved specifically into a demonstration of the challenges of training an Artificial Neural Network and how the challenges may be overcome
• finally in part 2, we covered some of the major components of an Artificial Neural Network, the major options available and where each may be suited to be used

PART THREE

(Video of PART 3 is at the top of this page)

In PART 3, we

• deep dive into the process of training an Artificial Neural Network step-by-step
• detailed the processes in the normal and customary technical way
• re-described the processes in a layman's term using a food supply chain as an allegory to represent an Artificial Neural Network (This could be beneficial to the reader)
• contrasted plain vanilla AI with something the likes of which the students may encounter in the real world, namely the application of Recurring Neural Network in a stock pricing solution
• the lecture ended examining why AI is now a key factor determining business competitiveness and how it is necessary for nation-building

ADDITIONAL MATERIAL

In contrast, here is an introductory lecture to Deep Learning from MIT. 

It would also be immensely beneficial to consider how Andrew Ng spoke about real-world AI and Deep Learning. Andrew is a prominent name in the world of AI who co-founded Google Brain, is at the time of writing Chief Scientist at Baidu, and an adjunct Professor at Standford. Andrew Ng grew up in Singapore. 

#fintech #artificialintelligence #digitaltransformation