Course Details
Course Details
What You'll Learn
Topic 1: Introduction to Reinforcement Learning
What is Reinforcement Learning (RL)?
Markov Decision Process (MDP) and RL
Applications of RL
RL Algorithms Classifications
Topic 2: OpenAI Gym
What is OpenAI Gym
Install OpenAI Gym
OpenAI Gym Operations
Topic 3: Value Based Q-Learning
What is Q-Learning
Q Value and Q-Table
Bellman Equation
Q-Learning Algorithm
Epsilon Greedy Explore-Exploit Strategy
On-Policy vs Off-Policy Learning
What is SARSA?
SARSA Algorithm
Topic 4: Policy-Based Learning
Policy Based Methods
Policy Gradient Algorithm
Implementation of Policy Gradient Algorithm
Topic 5: Overview of Advanced RL Algorithms
Limitation of Value and Policy-Based Learnings
Actor-Critic Algorithms
Deep Reinforcement Algorithms
Topic 6: Model-Based Learning
What is Model-Based Learnings
Model-Based Q-Learning Algorithms
Assessment
- Written Exam
- Practical Exam
Course Info
Minimum Entry Requirement
Knowledge and Skills
- Able to operate using computer functions with minimum Computer Literacy Level 2 based on ICAS Computer Skills Assessment Framework
- Minimum 3 GCE ‘O’ Levels Passes including English or WPL Level 5 (Average of Reading, Listening, Speaking & Writing Scores)
Attitude
- Positive Learning Attitude
- Enthusiastic Learner
Experience
- Minimum of 1 year of working experience.
- Minimum 18 years old
Minimum Software/Hardware Requirement
Software:
You will need the following for this course:
- Python 3 – free
- Visual Studio Code – free
- Google Colab – free, runs in the browser with a Google account
Hardware: Windows PC or Mac laptop
Funding Eligibility Criteria
Self-sponsored learners (individuals)
- Singapore Citizens and Singapore Permanent Residents aged 21 and above.
- Up to 50% course fee funding. Singapore Citizens aged 40 and above receive up to 70% under the Mid-career Enhanced Subsidy (MCES).
Company-sponsored learners
- Singapore Citizens and Singapore Permanent Residents who are direct employees of the sponsoring company.
- Up to 50% course fee funding, or up to 70% for employees of SMEs (Enhanced Training Support for SMEs) and for Singapore Citizens aged 40 and above. An SME has no more than 200 employees or an annual turnover of no more than $100 million.
- Employers can also claim Absentee Payroll funding for employees who attend during working hours.
- Eligible companies can offset their out-of-pocket fees with SkillsFuture Enterprise Credit (SFEC).
Conditions for all funded learners
- Attendance is taken digitally with Singpass for every session, including the assessment.
- You must complete the course and pass all assessments (assessed as Competent).
- If these conditions are not met, the funding is clawed back and the learner or employer pays the full course fee.
- 9% GST is charged on the full course fee before funding.
- Promo and discount codes cannot be used on funded courses.
- Any balance after funding can be paid in cash, with PSEA (Singapore Citizens with a PSEA), or partly reimbursed by UTAP for NTUC members.
How to Apply SkillsFuture Enterprise Credit (SFEC)
SFEC helps companies pay for employee training. The current SFEC ends on 30 November 2026, and a redesigned SFEC starts on 1 December 2026.
Current SFEC: classes ending on or before 30 November 2026
- Eligible companies received a one-off $10,000 credit, notified by letter to their Corppass administrator. Eligibility generally required at least 3 Singapore Citizen or PR employees and Skills Development Levy contributions.
- SFEC covers up to 90% of the company's out-of-pocket course fee after other subsidies.
- Check your company's SFEC eligibility and balance on EnterpriseSG.
- Register your Singapore Citizen or PR employee for a class that ends by 30 November 2026, as company-sponsored, and pay the course fee after funding.
- Your employee attends and completes the course. We submit the attendance and assessment results to SWDA.
- No separate claim is needed. SFEC is calculated automatically once the course fee grant is approved and is paid quarterly to your company's IRAS GIRO or PayNow Corporate account.
Any unused balance of the current SFEC expires on 30 November 2026.
Redesigned SFEC from 1 December 2026
Companies with at least 3 resident employees receive a fresh $10,000 in an online SFEC wallet that offsets out-of-pocket fees upfront at enrolment, instead of reimbursing them afterwards. SWDA has not yet published the support rates and claim steps; see OurWorkforce or our SFEC guide.
How to Apply UTAP (NTUC Members)
NTUC members can claim 50% of the unfunded course fee under the Union Training Assistance Programme (UTAP), capped at $250 a year (aged below 40) or $500 a year (aged 40 and above). The cap is shared with UTAP claims for AI tool subscriptions.
Eligibility: you must be an NTUC member paying full membership fees, with your membership paid up before and during the course and when you claim. The course must not be fully paid for by your employer or another source.
- Make sure your NTUC membership is paid up before the course starts.
- Register for this course with us and pay the course fee after funding.
- Attend at least 75% of the course and sit all assessments.
- Within 6 months after the course ends, go to the NTUC U Portal, open UTAP under E-services, log in with Singpass and choose Course Fee Claim.
- Enter the nett course fee you paid (after all subsidies and SkillsFuture Credit, including GST), and upload your invoice or receipt and your certificate or result slip (PDF, JPG or PNG, up to 5MB each).
- NTUC usually pays the claim within 4 weeks of submission.
Read more in our UTAP guide.
How to Apply PSEA
Singapore Citizens with a Post-Secondary Education Account (PSEA), which stays open until around the year you turn 31, can use it to pay the course fee after funding for PSEA-approved courses. You can also use a sibling's PSEA after your own has been used up.
- Check that this course shows the PSEA tag on its MySkillsFuture page.
- Check your PSEA balance on MOE's Student Finance System with Singpass.
- Download and complete the PSEA withdrawal form, then submit it to us.
- We submit the withdrawal request to MOE. You (and your sibling, if you are using their account) authorise it in the Student Finance System. Nothing is withdrawn without this authorisation.
- MOE pays us directly. If you have already paid, we refund the amount covered by PSEA.
More details: MOE PSEA overview.
Appeal Process
If you disagree with an assessment decision, you have the right to appeal.
- When the assessor gives you feedback, say whether you agree with the outcome. If you agree, you and the assessor sign the Assessment Summary Record; if you disagree, do not sign it.
- Discuss the decision with the assessor or the assessment manager first.
- If you are still not satisfied, tell the assessor you intend to appeal. The assessor records this in the feedback section of the Assessment Summary Record and informs the assessment manager.
- Lodge your appeal within 7 days, giving your reasons, using the Candidate Appeal Form. The assessor can help you write and lodge it.
- An Assessment Appeal Panel reviews the information from you and the assessor and makes a decision. A record of the appeal and its outcome is kept.
- You will be told the outcome within 2 weeks of lodging the appeal. The panel's decision is final.
Job Roles
Job Roles
- Machine Learning Engineer
- Robotics Engineer
- Game Developer (AI-focused)
- AI Research Scientist
- Data Scientist (branching into RL)
- Autonomous Systems Developer
- Simulation Engineer (using RL)
- Optimization Specialist
- AI Product Manager (oversight on RL projects)
- Control Systems Engineer (using RL)
- Finance Quant (using RL for trading strategies)
- NLP Engineer (using RL for certain applications)
- Recommendation System Developer (using RL)
- AI Solutions Architect
- Drone Algorithm Developer.
Trainers
Trainers
Dr. Alfred Ang's path to reinforcement learning runs through physics, electrical engineering and robotics. He holds a B.Sc (Hons) in Physics from NUS, where he topped his cohort, an MSc in Electrical and Electronic Engineering from NTU, a PhD in Electrical Engineering from NUS and a Graduate Diploma in Intelligent Robotics Systems from NUS-ISS. He worked in the defence, electronics and semiconductor industries, published more than 20 peer-reviewed papers and co-invented more than 20 US patents, and founded Tertiary Robotics in 2015 to supply Arduino, Raspberry Pi, Micro:bit and robotics products. His interests include machine learning, deep learning, artificial intelligence and robotics, and he has completed an Executive Programme in Algorithmic Trading from QuantInsti. Today he is Managing Director, CIO, CTO and Chief Instructional Designer of Tertiary Infotech Academy Pte Ltd, a full-stack AI developer and Agentic AI solution architect, a Microsoft Certified Data Scientist (DP-100) and AI Engineer Associate (AI-102), and an IBM certified instructor for the AI Practitioners course. His 300+ public GitHub repositories include interactive Python and Streamlit labs for beginners in data mining and machine learning, a stock analysis dashboard with an AI agent that recommends buy, sell or hold decisions with its reasoning, and interactive explainers on quantum computing and mathematics. He is also an ACTA-certified trainer and DACE-certified curriculum developer, and enjoys translating mathematically rich topics into accessible, practical lessons for working professionals, as his interactive explainers on mathematics and physics show.
In this CASL course, Dr. Ang makes reinforcement learning approachable for beginners by building intuition first and code second. Learners start with what reinforcement learning is, how Markov Decision Processes frame an agent interacting with its environment, where RL is applied and how its algorithms are classified. They then set up OpenAI Gym and learn its core operations before implementing value-based methods: Q-values and Q-tables, the Bellman equation, the Q-learning algorithm, epsilon-greedy exploration and the difference between on-policy and off-policy learning through SARSA. The course moves on to policy-based learning with a hands-on policy gradient implementation, an overview of actor-critic and deep reinforcement learning algorithms, and model-based Q-learning. His robotics and engineering background lets him relate abstract ideas such as reward design, exploration and control to tangible examples like robots learning to move or balance, and his trading studies inform discussion of sequential decision-making under uncertainty. Teaching is hands-on in Python, with learners training and observing their own agents in simulated environments, so they leave able to implement core RL algorithms and judge when RL is the right tool. He also explains practical issues that beginners often meet, such as choosing learning rates and discount factors, balancing exploration against exploitation, and recognising when an agent has stopped improving. Learners visualise rewards and learning curves to understand what their agents are doing, and discuss how the same principles extend to robotics, operations and game-playing applications. To see his machine learning labs and projects, view his GitHub portfolio at https://github.com/alfredang and connect with him on LinkedIn at https://www.linkedin.com/in/angchewhoe/
Tan Woei Ming is an AI engineer and data scientist with over 15 years of experience in machine learning, deep learning and intelligent automation. He has led AI projects in predictive analytics, image recognition and process optimisation within the semiconductor and manufacturing industries. His expertise includes Python, PyTorch, TensorFlow and reinforcement learning, where he applies computational models to improve decision-making and automation in complex systems. He holds a Master's in Intelligent Systems from the National University of Singapore and a First-Class Honours in Electrical and Electronic Engineering from NTU.
In CASL - Practical Reinforcement Learning for Beginners, Woei Ming introduces the foundations of reinforcement learning through hands-on coding and simulation. Learners cover Markov decision processes and RL applications, set up OpenAI Gym, and implement value-based Q-learning with the Bellman equation and epsilon-greedy exploration. They go on to compare on-policy and off-policy learning with SARSA, implement policy gradient methods, and survey actor-critic, deep reinforcement learning and model-based approaches. By combining theory with practical Python exercises, he helps learners build intuition for how agents learn through interaction, preparing them to apply these ideas to automation and optimisation problems.
Solomon Soh is a data scientist and AI trainer with extensive experience in reinforcement learning, deep learning and optimisation. At Workforce Optimizer he spearheaded R&D in job-shop reinforcement learning, developing solutions that improved operational efficiency by 15% and reduced staffing costs through discrete optimisation. At IBM Singapore he supervised 24 machine learning and deep learning projects, coaching teams on methodologies, feature engineering and model deployment, and earned a 96% learner satisfaction rating. He holds a double degree in Economics and Psychology from Singapore Management University (Summa Cum Laude), an MBA and a Master's in Financial Engineering.
CASL Practical Reinforcement Learning for Beginners benefits directly from his industrial RL work. Learners are introduced to reinforcement learning, Markov decision processes and the main classes of RL algorithms, set up OpenAI Gym, and implement value-based Q-learning with the Bellman equation and an epsilon-greedy strategy, followed by SARSA, policy gradient methods, actor-critic and deep reinforcement learning, and model-based learning. Solomon emphasises practical applications of RL in scheduling, forecasting and resource optimisation, drawing on real projects to make abstract ideas concrete. Participants leave with a solid conceptual understanding and the hands-on Python skills to start applying reinforcement learning to their own problems.
Quah Chee Yong is an ACLP-certified trainer and data science professional with strong expertise in machine learning, natural language processing and AI systems. As AI Solutions Lead at AiDeal Scan, he developed recommender systems and NLP-driven search engines, while at GoWild Singapore he led the data science team to build analytics platforms and chatbot solutions powered by reinforcement learning. He has also served as Data Science Training Lead for SAP and Temasek Polytechnic programmes under IMDA, delivering practical AI and machine learning training for professionals.
In CASL Practical Reinforcement Learning for Beginners, he helps learners understand reinforcement learning through applied case studies, with examples drawn from customer analytics, recommender systems and chatbots. Participants cover what reinforcement learning is, Markov decision processes and the main families of algorithms, then set up and operate OpenAI Gym environments. They implement value-based Q-learning, including Q-tables, the Bellman equation and the epsilon-greedy explore-exploit strategy, compare on-policy and off-policy learning with SARSA, and move on to policy-based learning. By combining commercial project experience with training expertise, he equips beginners to apply these techniques with confidence.
Dr Alvin Ang is an ACLP-certified trainer with a PhD in Operations Research from Nanyang Technological University and more than a decade of experience in AI, optimisation and applied machine learning. He has taught at NTU, SUSS, Curtin University and SP Jain School of Global Management, and served as an IBM Data Science Instructor. His certifications include TensorFlow, machine learning with Python and deep learning with Keras, supported by extensive experience in AI model design and deployment.
In CASL Practical Reinforcement Learning for Beginners, Dr Ang introduces reinforcement learning fundamentals, an area where his operations research and optimisation background is a natural fit. Learners cover Markov decision processes and the main families of RL algorithms, set up and operate OpenAI Gym, and implement value-based Q-learning with the Bellman equation and an epsilon-greedy strategy, followed by SARSA, policy gradient methods, actor-critic and deep reinforcement learning, and model-based learning. He explains core ideas such as exploration versus exploitation and reward structures, and emphasises practical implementation in Python. Blending academic rigour with coding exercises, he ensures participants leave able to design and evaluate RL models in applied settings.
Review
Customer Reviews (5)
- might recommend Review by Course Participant/Trainee
-
. (Posted on 3/26/2023)1. Do you find the course meet your expectation? 2. Do you find the trainer knowledgeable in this subject? 3. How do you find the training environment - recommended Review by Course Participant/Trainee
-
Informative and it's great to learn practically. Would have been nice to learn how to create a gym for a new unique scenario1. Do you find the course meet your expectation? 2. Do you find the trainer knowledgeable in this subject? 3. How do you find the training environment
I recorded the links I used to install the packages for scripts in an m1 env. Let me know if you need it
Bash files to install packages and environment (Posted on 7/31/2022) - will recommend Review by Course Participant/Trainee
-
Tertiary Courses has put much effort in preparing the material. There is a well balance between the theory and coding. Having said that, RL is also a pretty tough subject. Recommend to split two training days into non consecutive days or can considering to increase 2-day training to 3-day training. (Posted on 6/17/2022)1. Do you find the course meet your expectation? 2. Do you find the trainer knowledgeable in this subject? 3. How do you find the training environment - will recommend Review by Course Participant/Trainee
-
. (Posted on 6/17/2022)1. Do you find the course meet your expectation? 2. Do you find the trainer knowledgeable in this subject? 3. How do you find the training environment - will recomendation Review by Course Participant/Trainee
-
More applications to commercial domains1. Do you find the course meet your expectation? 2. Do you find the trainer knowledgeable in this subject? 3. How do you find the training environment
Good attempt for a difficult topic, keep it up! (Posted on 1/23/2022)
Write Your Own Review
- Recommended Courses
- Related Courses