Data Science Weekend

Data Science Center LST FMIPA UI

Come and Join

Weekend Course 1:

Generative Adversarial Networks In-Depth

Participants : Data Scientists, AI Engineers and Researchers (Max 30)
Prices : Students (IDR 1mill, Professional IDR 5mill)
Duration : 4 Weeks (Saturday 09.00 – 16.00)
Method : Fundamental, Conceptual and Practical Approaches
Guess Speaker : Risman Adnan, Samsung R&D Institute Indonesia

Tagline:

Generative models like GANs become hot topic in deep learning technology with various industrial use cases on vision, speech and text generations. In this weekend course, we will cover fundamental and conceptual understanding of GANs and equip participants with practical skills to train image, text and video datasets with state of art GAN models.

Syllabus:

Week 1: Fundamental of Generative Adversarial Networks
Week 2: Minimax and Wasserstein GAN Formulations
Week 3: Various GAN Evaluation Methods
Week 4: Training Stable GANs for Image, Text and Video Dataset.

Course Materials:

  • Lecture Slides
  • Reusable PyTorch Codes
  • Home Work Problems

Weekend Course 2:

Machine Learning with Kernels

Participants : Data Scientists, AI Engineers and Researchers (Max 30)
Prices : Students (IDR 1mill, Professional IDR 5mill)
Duration : 4 Weeks (Saturday 09.00 – 16.00)
Method : Fundamental, Conceptual and Practical Approaches
Guess Speaker : Risman Adnan, Samsung R&D Institute Indonesia

Tagline:

Kernel based models are widely use machine learning technique for both discriminative and generative tasks. In this course we will introduce Reproducible Kernel Hilbert space (RHKS) as key concept to understand and implement various kernel based models. Based on RHKS understanding, we will show how to use it for various machine learning discriminative tasks. We will also introduce concept of Maximum Mean Discrepancy to train Generative Adversarial Networks.

Syllabus:

Week 1: Introduction to Machine Learning with Kernels
Week 2: Reproducible Kernel Hilbert Space in Machine Learning
Week 3: Represent and Compare Probabilities with Kernels
Week 4: MMD for Training Generative Adversarial Networks

Course Materials:

  • Lecture Slides
  • Reusable PyTorch Codes
  • Home Work Problems

Weekend Course 3:

Computational Optimal Transport

Participants : Data Scientists, AI Engineers and Researchers (Max 30)
Prices : Students (IDR 1mill, Professional IDR 5mill)
Duration : 5 Weeks (Saturday 09.00 – 16.00)
Method : Fundamental, Conceptual and Practical Approaches
Guess Speaker : Risman Adnan, Samsung R&D Institute Indonesia

Tagline:

Optimal Transport (OT) Theory opens new paradigm on both discriminative and generative machine learning tasks. Our focus is on the recent wave of efficient OT algorithms that have helped translate attractive theoretical properties onto elegant and scalable tools for a wide variety of machine learning and deep learning applications. We will start from basic theoretical and algorithmic foundations, then introduce various regularisation techniques for optimal transport divergence. We will also introduce how to use regularized OT to train Generative Adversarial Networks.

Syllabus:

Week 1: Theoretical and Algorithmic Foundations
Week 2: Optimal Transport Applications on Data Science
Week 3: Entropic Regularization and Semi-Discrete Methods
Week 4: Gradient Flows and Gromov Wasserstein
Week 5: Density Fitting and Generative Adversarial Networks

Course Materials:

  • Lecture Slides
  • Reusable Python Codes
  • Home Work Problems