COM503: Introduction

COM503: Deep Learning is a centerpiece of the SIAI's MBA AI/BigData (AI/Finance) program, together with COM502: Machine Learning and COM504: Reinforcement Learning.

The course assumes prior knowledge in regression analysis in following courses along with important understanding of key concepts in computational efficiency (discussed in COM501: Scientific Programming) and basic understanding of computational modeling (discussed in COM502: Machine Learning and BUS501: AI in Digital Marketing )

• STA501: Data-based Decision Making
• STA502: Math & Stat for MBA I
• STA503: Math & Stat for MBA II

This course is dedicated to artificial neural network models and its extensions. At first, the course covers Autoencoder, the model of which is a graph model version of nested Factor Analysis. The Lecture 3 detours to MCMC and Bayesian tools to learn Boltzmann machine in Lecture 4.

After all the basic tools are processed, the course covers CNN, RNN, and generative models, all of which are variations of basic Deep Neural Networks for specific purposes and data structures.

Final term paper continues its discussion from COM502: Machine Learning. As for COM502: Machine Learning with BUS501: AI in Digital Marketing, BUS502: AI in Business Cases is a business case course taught in parallel.

Course topics

.

Course schedule

.

Class discussion

The school strongly encourages in-class discussion, which covers 10% of the final grading. Students are given a discussion forum in all Moodle course pages, and various types of participations will be added to the final mark.

Plz note that only productive discussion activities will be positively marked. If any abusing is found, the student or the group of students will lose all 10% mark.

Final Examination

There will be a take-home final exam with 48 hours, which is usually provided on a friday 2 weeks after the last class.

Refer the sample exam questions for more details.

Grading and evaluation

Final marks will be broken down as below.

• 80%: Final examination
• 10%: Problem set submission (not graded - submission w/ reasonable amonnt work is sufficient)
• 10%: Forum discussion

For business track students, instead of final examination and problem sets, there will be an essay requirement covering 90% of the grading.

Essays are graded based on following criteria.

• Critical thinking
• Effective use of in-class knowledge
• Problem detection and solving in data scientific cases

All essays must not exceed 3,000 words.

Language

All contents must be submitted in English, although the school does not refrain students relying on AI tools for translation.