Select Page

9 – Reinforcement learning

Webinar 9: Reinforcement Learning

The session is designed to provide a comprehensive overview of RL, understanding its core concepts, key algorithms, practical applications, and the challenges it faces. We begin by defining RL and distinguishing it from other types of machine learning, namely supervised and unsupervised learning. The session includes a detailed explanation of Markov Decision Processes (MDPs), which serve as the mathematical framework for RL, highlighting states, actions, transition probabilities, and rewards. The concepts of policies, value functions, and the exploration-exploitation trade-off are also examined to give participants a robust understanding of how RL agents learn and make decisions. The core of the session focuses on key RL algorithms, starting with classical methods such as Dynamic Programming, including policy iteration and value iteration. We then explore Monte Carlo methods and Temporal Difference learning, with specific emphasis on Q-Learning and SARSA. Advanced methods such as Deep Q-Networks (DQN) method is also covered, illustrating how deep learning integrates with RL to solve complex problems. 

Abebaw Degu is an engineer, teacher, and researcher. He holds BSc in software engineering and MSc in Computer science and engineering from Adama science and Technology University, and a PhD in Artificial Intelligence from Euromed University of Fes. Abebaw’s research focuses on reinforcement learning and transfer learning, particularly in the areas of sequential decision-making and optimization. He is actively engaged in innovative research and has published papers and presented his work in international conference in these fields.

 

 

 

 

 

8 – Learning Evaluation

Webinar 8: Learning Evaluation

Learning evaluation is crucial in assessing the performance and reliability of machine learning models. Trainees will grasp the importance of separating data into train, validation, and test sets to ensure unbiased evaluation. Performance metrics for classifiers and regressors, including error, precision, recall, and confusion matrix, will be elucidated to measure model effectiveness. Additionally, techniques such as cross-validation aid in robust performance evaluation. Trainees will learn to tune model parameters using validation sets and gain insights into understanding model behaviors, pitfalls, and implications of decisions. This session serves as a primer for trainees to navigate the intricacies of model evaluation, fostering informed decision-making in machine learning applications.

Hajar Alhijailan is an Assistant Professor of Computer Science at King Saud University in Riyadh, Saudi Arabia. She received her Ph.D. in Computer Science from the University of Liverpool in Artificial Intelligence. She taught AI courses for students in Computer Science Department at King Saud University.

 

 

 

 

 

7 – AI-Search

Webinar 7: AI-SEARCH

AI-search is a foundational aspect of problem-solving, essential in modern AI applications. This session introduces trainees to key concepts such as state space representation, uninformed and heuristic graph search algorithms, and complexities analysis. Trainees will learn to navigate problem spaces efficiently through algorithms like breadth-first search, depth-first search, and A* search, balancing optimality and computational efficiency. Understanding the space and time complexities of these algorithms enables effective problem-solving strategy selection. Additionally, the session covers strategies for two-player adversarial games, including minimax search and alpha-beta pruning, enhancing trainees’ ability to compete in dynamic environments. Mastering these concepts equips trainees with essential skills for tackling diverse real-world challenges.

Hajar Alhijailan is an Assistant Professor of Computer Science at King Saud University in Riyadh, Saudi Arabia. She received her Ph.D. in Computer Science from the University of Liverpool in Artificial Intelligence. She taught AI courses for students in Computer Science Department at King Saud University.

 

 

 

 

6 – Supervised learning

Webinar 6: Supervised Learning

In this session, we will explore the fundamentals of supervised learning methods in machine learning, including Linear Regression, Logistic Regression, Decision Trees, Support Vector Machines (SVM), and K-Nearest Neighbors (KNN). Each method will be illustrated with real-world examples to demonstrate their practical applications. Attendees will gain a solid understanding of how these algorithms work, their use cases, and how to implement them effectively in Python. By the end of the session, participants will be equipped with the knowledge to apply these supervised learning techniques to solve various predictive modeling problems.

Dr. Nouf AlShenaifi is a lecturer, researcher, and PhD candidate at King Saud University, specializing in Artificial Intelligence, machine learning, and data science. Her research focuses on Natural Language Processing methods. She actively engages in innovative research, with work published in prestigious journals and presented at international conferences, reflecting her commitment to advancing the field of Artificial Intelligence. 

 

 

 

5 – Deployed deep generative models

Webinar 5: Deployed Deep Generative Models

Deployed deep generative models have transformed AI applications, particularly in generating images and text. This session covers understanding these applications, ensuring data cleanliness, and identifying suitable tasks. It explores deep image generative models like DALL-E, Midjourney, and Stable Diffusion, highlighting their use in art and design, and challenges like bias and high computational costs. It also delves into large language models like ChatGPT and Bard, explaining their transformer architectures and uses in customer support and content generation, while addressing issues like potential biases and significant computational requirements.

Sarah Alotaibi is an Assistant Professor in the Department of Computer Science at King Saud University. She holds a B.Sc. and M.Sc. in Computer Science from King Saud University and a Ph.D. in Computer Vision from the University of York in the United Kingdom. Her research focuses on computer vision and machine learning, with an interest in deep learning with statistical and appearance modeling, face modeling, reflectance analysis, and inverse rendering. Dr. Alotaibi has published numerous papers in these areas.

 

 

 

4 – Working with Data

Webinar 4: Working With Data

This session covers critical data preprocessing steps essential for machine learning model performance and accuracy. Trainees will learn techniques for handling missing values, such as imputation and flagging, and encoding categorical and real-valued data into numerical formats. The session will include normalization and standardization to improve gradient-based algorithms’ convergence rates, and data preparation methods like one-hot encoding and representation. Understanding the hypothesis space and managing biases through cross-validation are key topics. The session will also cover partitioning data into training, validation, and test sets, and optimizing model performance through parameter tuning with grid or random search.

Sarah Alotaibi is an Assistant Professor in the Department of Computer Science at King Saud University. She holds a B.Sc. and M.Sc. in Computer Science from King Saud University and a Ph.D. in Computer Vision from the University of York in the United Kingdom. Her research focuses on computer vision and machine learning, with an interest in deep learning with statistical and appearance modeling, face modeling, reflectance analysis, and inverse rendering. Dr. Alotaibi has published numerous papers in these areas.