Course description

This course aims to give leaners a very direct and accessible introduction to the principles & foundations of fields of AI, Machine Learning (especially Deep Learning), and Robotics. More specifically:

  1. Equipping learners with a unified view of Machine Learning – the core of AI – via the TEFPA Framework (Task, Experience, Function space, Performance measure, Algorithm to search/optimize) proposed by COTAI. This allows learners to understand the many ML algorithms as special cases of a big picture.
  2. Providing learners with an intuitive understanding of the key concepts, ideas, and math foundations of AI/ML through our inspiring lectures and project-based curriculum.
  3. Guiding learners step-by-step how to best implement the discussed algorithms in core AI areas including visual and language understanding, and decision making. This enable learners to create exciting and practical Deep Learning applications utilizing pretrained models.

Besides, learners will also have chance to review and practice the following foundational knowledge & skills (through accompanied tutorial & lab sessions, homework assignments & projects):

  • Computer and Programming Fundamentals: Data Structure & Algorithms, Git & Version Control for Collaborative Dev, Python programming
  • Math Foundations: Abstract Vector Spaces, Linear Algebra, Matrix and Transformations, Multivariable Calculus, Optimizations, Probability & Statistics, Information Theory, Logic.
  • Data Engineering Tools: Data Web Crawler, Data Visualization.

Course outline

 Session Content
1 From intelligence & learning to AI & ML. The TEFPA framework (part 1)
2 Basic math: linear algebra, and abstract vector space
3 Applied math: probability, statistics, and information theory
Session Content
4 Representation issues: deep feedforwad network, feature extraction, and embedding coordinates
5 Evaluation issues: common loss functions
6 Search issues: gradient descent and its variations
Session Content
7 More on representation issues: CNNs for grid-like data
8 More on representation issues: RNNs for time-series-like data
9 More on search issues: overfitting, underfitting, and regularization
Session Content
10 Data acquisition, cleaning, annotation
11 Data exploration: visualization, statistics, imbalance, patterns, etc.
12 Lab lecture (Kaggle contests)
Session Content
13 Introduction to image classification: AlexNet and VGGNet
14 More on image classification: ResNet and MobileNet
15 Lab lecture
Session Content
16 Introduction to basic modules: word2vec, LSTM, and GRU
17 Introduction to basic modules: Transformer and attention mechanism
18 Lab lecture

 

Session Content
19 Decision making: classical planning & Reinforcement Learning
20 Contextual bandits and interactive decision making
21 Lab lecture

 

Session Content
22 Practical methodology. Start an intra-class competition (final project)
23 Lab lecture
24 Lab lecture. Final project report and ranking

Course Execution:

  • 72 Hours, 8 Weeks, 3 sessions/week, 3 hours/ session.
  • Blended-Learning

Learning outcomes (Exit Competencies):

Upon completing this course, student should be able to:

  • Understand overview of AI/ML
  • Use applied Math, Python, Data engineering tools for AI/ML
  • Be knowledgeable of ML/DL Models
  • Unify machine learning algorithms into a big picture following TEFPA framework
  • Apply algorithms to build AI applications utilizing pretrained Deep Learning Models