Skip to main content
Institute of Advanced Technological and Commercial Studies

Glossary

Machine learning: learning the rule from examples

Machine learning builds a decision rule from examples rather than programming it explicitly. The resulting model is worth exactly what the data that trained it is worth. Its quality is measured on held-out data, never on the examples it learned.

From example to rule

Machine learning fits a model on annotated examples until it recovers the expected label, then measures its quality on data it has never seen. The result is judged on that held-out share, never on the training examples. The work concentrates on the data, on defining what is to be predicted, and on measuring the real effect once the model is in service.

  • The model learns a rule from annotated examples
  • Quality is measured on data held out for evaluation
  • The real effect is observed once the model is in service

The full life of a model

The Artificial Intelligence programme runs the full cycle: problem stated, dataset assembled, model fitted, deployment and monitoring. The Data Science programme approaches the same cycle through statistics, insisting on what the result allows you to assert and what it calls on you to verify.

  1. 01

    1 — State the question

    What is to be predicted, and the decision that prediction must enable.

  2. 02

    2 — Assemble the data

    Sources brought together, variables built, evaluation share held out before any training.

  3. 03

    3 — Fit and measure

    The model is fitted on the training share, then measured on the held-out share, and on it alone.

  4. 04

    4 — Deploy and monitor

    The model goes into production with its monitoring indicators and a periodic review of its results.

Programmes that take a model into service

2 catalogue programmes put “Machine learning” to work: 240 credits and 2400 taught hours in total. The official rule holds throughout — one credit stands for 25 hours of work, 10 of them taught. Every line below is recomputed from the programme page: level, duration, credits, taught volume and fees in Guinean francs appear exactly as filed in the official catalogue.

ProgrammeLevelDurationVolumeFees
Artificial Intelligence programme — Bac+5 levelMaster's degree2 years120 credits · 1200 taught hours7 800 000 GNF per year
Data Science programme — Bac+5 levelMaster's degree2 years120 credits · 1200 taught hours7 200 000 GNF per year

Your questions, our answers

What level of mathematics is needed?

The exact prerequisite appears on the programme page. It covers linear algebra, probability and descriptive statistics.

Do you work on African data?

Yes: the work is grounded in datasets from the sub-region and in real deployment cases.

Which professional qualification does this skill prepare?

From applied data and artificial intelligence technician up to data leadership functions, depending on the level targeted.

Machine learning: learning the rule from examples — Glossary | IHETC — IHETC