Three Regions of the Learning Landscape
A detailed look at what each programme covers, how it is taught, and what it asks of students.
Back to HomeHow the Programmes Are Taught
Assigned Readings
Each week builds from selected papers and textbook chapters. Students are expected to engage with them before the live session.
Live Weekly Session
Substantive discussion of the week's material with the instructor. Questions on implementation, theory, and conceptual clarification.
Projects With Feedback
Implementation and analysis projects are submitted and reviewed by the instructor. Written feedback is returned within one week.
Probabilistic Machine Learning
A structured 14-week programme that examines machine learning through a probabilistic framework. Students work through Bayesian inference, graphical models, variational methods, Monte Carlo approaches, and Gaussian processes — building each from scratch using NumPy and PyMC before reaching for higher-level abstractions. The focus on uncertainty quantification runs throughout: not just as a technique, but as a mode of thinking about model behaviour.
Topics Covered
- Bayesian inference: prior, likelihood, posterior
- Directed and undirected graphical models
- Variational inference and the ELBO
- Monte Carlo methods including MCMC and HMC
- Gaussian processes: regression and classification
- Uncertainty quantification in practical contexts
Solid probability and linear algebra background. Prior exposure to machine learning fundamentals. Comfort with Python and NumPy.
Weekly Process
Assigned reading distributed at week start. Study at your own pace over several days.
Live session mid-to-late week. Conceptual discussion, worked examples, and Q&A.
Implementation exercise assigned after the session. Code and analysis submitted by week's end.
Larger project milestones reviewed with written feedback by the instructor.
Student Projects
Project 1: Audio feature extraction and analysis pipeline. Implement and compare MFCC and spectrogram representations on a curated speech dataset.
Project 2: End-to-end ASR implementation using a modern architecture. Evaluate against established metrics and analyse failure modes.
Project 3: Text-to-speech system evaluation. Implement and assess a synthesis pipeline; document design choices and their acoustic consequences.
Applied Speech Processing
A 12-week programme addressing speech processing methods and their application. The curriculum moves from audio signal fundamentals through traditional recognition approaches to modern end-to-end architectures and text-to-speech systems. Students work in PyTorch with established speech libraries, completing three substantial projects spanning recognition and synthesis. Instructor review of each project submission is included in the programme.
Topics Covered
- Audio signal processing: sampling, spectral analysis, features
- Traditional speech recognition: HMMs, acoustic modelling
- End-to-end architectures: CTC, attention-based approaches
- Text-to-speech systems and neural vocoding
- Evaluation methodology for speech tasks
Deep learning familiarity and practical experience with PyTorch or TensorFlow. Basic signal processing background is helpful but can be built during the programme's opening weeks.
Foundations of Optimization
A 10-week programme focused on optimization as applied in machine learning contexts. The curriculum addresses convex optimization fundamentals, gradient-based methods and their variants (SGD, Adam, and related approaches), second-order methods, constrained optimization, and non-convex topics directly relevant to neural network training. Theory and implementation are balanced throughout — the mathematical foundations are treated seriously, and implementation exercises consolidate understanding.
Topics Covered
- Convex sets, functions, and optimization problems
- First-order methods: gradient descent and variants
- Stochastic and mini-batch optimization
- Second-order methods and their approximations
- Constrained optimization and Lagrangian duality
- Non-convex landscapes in neural network training
Calculus (multivariable) and linear algebra. Suitable as preparation for the Probabilistic ML programme, or for practitioners who want deeper theoretical grounding in how learning algorithms actually converge.
Who This Programme Serves
Foundations of Optimization is appropriate for two distinct groups:
Practitioners building models who want to understand why their training behaves as it does — why certain learning rate schedules work, why adaptive methods converge differently, what loss landscape geometry implies.
Students preparing for the Probabilistic ML programme who need to build or consolidate their mathematical foundations first.
Programme Comparison
Use this table to understand what distinguishes each programme and which is the right starting point.
| Attribute | Prob. ML | Speech Proc. | Optimization |
|---|---|---|---|
| Duration | 14 weeks | 12 weeks | 10 weeks |
| Price (MYR) | 4,100 | 3,400 | 2,700 |
| Math intensity | High | Moderate–High | High |
| Recommended starting point | — | ||
| Instructor project review | |||
| Weekly live session | |||
| Deep learning prerequisite | Helpful | Required | Not required |
| Best for | ML engineers wanting probabilistic reasoning | Deep learning practitioners building audio systems | Anyone wanting mathematical training foundations |
Standards That Apply Throughout
Data Privacy
Student data handled per Malaysia's PDPA 2010. No data shared with third parties without explicit consent.
Quality Assurance
Materials reviewed before each intake. Instructor performance assessed post-programme against qualitative outcomes.
Student Support
Administrative and logistical queries responded to promptly. Technical issues escalated directly, not through a ticketing queue.
Transparent Expectations
Prerequisites, workload, and programme scope communicated clearly before enrolment. No surprises about what is expected of students.
Programme Fees
Not Sure Which Programme to Start With?
Send us a brief note about your background and interests. We will tell you which programme fits your current level and what, if anything, to consolidate first.
Get Placement Advice