Machine Learning 1 [Odd Semester 2026/2027 Telyu] - Introduction to Machine Learning

Item: Machine Learning 1 [Odd Semester 2026/2027 Telyu] - Introduction to Machine Learning
Rating: 4
Author: PostSingularity Institute

by Machine Learning Indonesia · View the channel on PSI

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQMMachine learning

🎙 Machine Learning Indonesia 👥 3K 📅 September 20, 2026 ⏱ 61 min 👁 25 📄 tutorial 🧭 2026-09-20

Available in: English (current) Français

Keywords

machine learningFCP frameworksupervised learningunsupervised learningreinforcement learning

Summary

This is the first lecture of a Machine Learning course at Telkom University for the Odd Semester 2026/2027, delivered by an industry practitioner with experience at Microsoft, Samsung Research and Paragon. The instructor introduces the FCP learning framework — Fundamental, Conceptual, Practical — as the guiding methodology of the course, arguing that engineers must balance mathematical understanding, conceptual insight and practical implementation skills. He defines machine learning as techniques that learn patterns in data scientifically, and explains that a problem qualifies as a machine learning problem only when the target function is unknown and no explicit mathematical rules can be written, despite having sufficient data. Examples include handwritten digit recognition and bank credit approval decisions. The lecture formalizes the ML paradigm as a hypothesis set combined with a learning algorithm that selects a final hypothesis approximating the unknown target function. Three learning paradigms are compared: supervised, unsupervised (including semi-supervised) and reinforcement learning, with a chess example illustrating sequential decision-making under uncertainty. The instructor recommends two reference textbooks — PRML by Bishop and Learning from Data by Abu-Mostafa — and suggests practical tools (Kaggle, GitHub, Orange, scikit-learn, PyTorch, Google Colab). Throughout, the lecturer insists on mastering fundamental mathematics rather than merely generating code with AI tools.

205 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture provides strong pedagogical value: the FCP (Fundamental, Conceptual, Practical) framework offers a clear, actionable lens for approaching ML studies, and the repeated insistence on pairing mathematical rigour with hands-on practice is well argued. The argumentation for what constitutes a machine learning problem — an unknown mapping f with sufficient data and no available mathematical rule — is explained through two memorable examples (handwritten digits, bank credit approval). The instructor supports each claim with concrete industry experience and references to authoritative textbooks and scientists (Feynman, Abu-Mostafa). The structure is coherent: logistics → framework → definition → paradigms, and each concept is connected to both theory and practice. Some claims (e.g., that the future job market will be polarized between high-tech and agriculture) are presented as personal conviction rather than evidenced analysis, and the digressions on student motivation, though valuable pedagogically, occasionally dilute the informational density.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is acceptable for an introductory course: the instructor correctly describes fundamental ML concepts (hypothesis set, learning algorithm, supervised/unsupervised/reinforcement paradigms) and recommends two canonical textbooks with verifiable references. The mention of statistical learning theory (that a sample, not the whole data set, suffices) aligns with established results such as the Vapnik–Chervonenkis framework. Course material is openly shared on GitHub, which adds transparency. The title accurately mirrors the content: this is indeed an introduction to machine learning, structured as a first lecture of a semester. No discordant or unsupported factual claims were flagged; the main limitation is the absence of numerical experiments or published data, but this is appropriate for a lecture of this nature.

277 words

Title / Content Match

The title accurately reflects the content: this is a first introductory lecture of a machine learning course for the odd semester 2026/2027 at Telkom University.

Quality & Reliability

7/10

Expert instructor with an industrial background (Microsoft, Samsung Research, Paragon). Presents standard machine learning concepts (hypothesis set, learning algorithm, paradigms) correctly and cites recognized textbooks. The lecture is well-structured but introductory: no empirical data or experiments, and some claims rest on personal conviction rather than evidence.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

TeachingMLDL — Course material repository — Official repository of the course material, referenced in the video description.
RantAI · MLVR — Machine Learning via Rust — Promoted in the description as a detailed guide for learning ML with Rust.
RantAI Academy — Website of the educational platform associated with the course.
Rustindo — Telegram community — Community channel mentioned in the description for support.
RantAI on LinkedIn — LinkedIn page of the community backing the course.
Pattern Recognition and Machine Learning (C. Bishop) — Textbook recommended by the instructor, licensed by Microsoft and freely accessible.
Learning From Data (Y. Abu-Mostafa) — Textbook recommended by the instructor, by a Caltech professor described as a Feynman award recipient.

Concurring Sources

Pattern Recognition and Machine Learning (C. Bishop) — Aligns with the course content on supervised learning and linear models presented in the lecture.
Learning From Data (Y. Abu-Mostafa) — Covers the same ML paradigm (hypothesis set, learning algorithm, VC theory) discussed in the lecture.

Contribution & Novelties

The lecture’s added value lies in its pedagogical framework: the FCP triad offers a simple mental model students can apply to any ML topic, and the clear criterion distinguishing ML problems from rule-based problems (unknown target function + sufficient data) helps avoid conceptual confusion. The industrial perspective, urging students to implement ML in C++/C/Rust rather than Python alone, is a distinctive contribution rarely found in academic lectures. The hypothesis-set / learning-algorithm formalization provides a rigorous mental scaffolding for understanding any ML model, regardless of specific algorithm.

Pour aller plus loin :

Pattern Recognition and Machine Learning (PRML) — The main textbook recommended by the instructor; freely hosted by Microsoft, a standard reference in the ML field.
Machine learning — Encyclopedia overview covering the history, paradigms and applications of ML.
Supervised learning — Wikipedia article detailing the paradigm that occupies most of this course.
Reinforcement learning — Reference on the RL paradigm introduced in class, with the chess/sequential-decision illustration.
Learning From Data — Textbook by Yaser Abu-Mostafa (Caltech), also recommended in the lecture; its lectures are freely available on YouTube.

179 words

Radar Profile

The radar shows balanced scores across most axes (information quantity, quality and reliability all at 7/10), with the expected lower score on technical depth (5/10) for an introductory lecture. This profile indicates a solid, trustworthy educational resource that prioritizes conceptual clarity over technical progression.

Reliability 7/10