Intelligent Physics · Week 01 of 14

A century of physics has been theory-first: write the Hamiltonian, solve for the state. Machine learning inverts the order — start from data, let structure emerge.

Foundations No Prior ML Required Week 1

Objectives

By the end of this week, you will be able to:

  • State a precise, working definition of machine learning and explain how it differs from a traditional physics model
  • Distinguish supervised learning from unsupervised learning, and identify which one applies to a given problem
  • Define and correctly use the vocabulary: feature, label, model, training, prediction, dataset
  • Walk through one supervised example (predicting a property) and one unsupervised example (finding groups) by hand, on a tiny dataset
  • Take a position, with concrete physical reasoning, on where machine learning can and cannot help in your own area of materials physics

Prerequisites

None from machine learning. This week assumes only:

  • Familiarity with basic physical quantities relevant to materials (composition, structure, an example property such as formation energy or band gap)
  • Comfort reading a small table of numbers

No programming and no statistics background is required for this week. Both begin in Week 2.

This week's pages

Each week of this module is organized into four linked pages, reached from this presentation post. Work through them in order.

01 · Course Lesson The full lesson: definitions, two worked examples (one supervised, one unsupervised), and the data-vs-theory framing. 02 · Directed Work Travaux Dirigรฉs Guided pencil-and-paper problems on a small materials dataset, with full step-by-step solutions. 03 · Practical Work Travaux Pratiques No coding this week — instead, a structured preparation task for the Week 1 in-class debate. 04 · Quiz Self-Check Quiz Ten short questions to verify you can use this week's vocabulary correctly before Week 2.
How to use these four pages

Read the Lesson page first, in full, before attempting the Directed Work. The Directed Work problems reuse the exact dataset and vocabulary introduced in the Lesson, so they will not make sense out of order. The Quiz is meant to be attempted last, as a check, not as a first pass.

Looking ahead

Week 2 introduces the statistical tools (bias, variance, cross-validation) needed to judge whether a model has actually learned something useful, or merely memorized its training data. Everything in Week 2 builds directly on the vocabulary defined this week.

Intelligent Physics — Master's Module Week 2 →

INTELLIGENT PHYSICS: MACHINE LEARNING METHODS FOR MATERIALS SCIENCE — WEEK 1 OF 14