🤖 ML Types Explorer

Post 4 companion · Supervised · Unsupervised · Reinforcement Learning

Supervised Regression — Band Gap Prediction

A simple linear model trained on 5 DFT examples. Adjust the new material's features and see the predicted band gap.

Material	Lattice a (Å)	ΔEN	Eg DFT (eV)
ZnO	3.25	1.79	3.44
GaAs	5.65	0.40	1.42
Si	5.43	0.00	1.12
TiO₂	4.59	1.90	3.05
InP	5.87	0.11	1.34
❓ New	–	–	Predict →

Lattice constant a (Å) 4.0

Electronegativity diff ΔEN 1.0

Predicted Band Gap

–

Regression line (a vs Eg)

K-Means Clustering — Materials in Feature Space

30 materials plotted by two features (band gap vs lattice constant). Click "Cluster" to run K-Means and discover natural groups.

RL Agent — Synthesis Optimisation Game

The agent 🤖 explores a 5×5 grid of synthesis conditions (temperature × pressure). It gets +10 for stable materials (🟢), −5 for unstable (🔴), −1 per step. Watch it learn to find the best conditions.

Steps

Total Reward

Best Found

–

Agent initialised. Press Step or Auto Run...

1. You want to predict whether a compound is a metal or insulator using DFT labels. Which ML type and sub-type?

2. In reinforcement learning, what is the "reward" in a materials science context?

3. PCA (Principal Component Analysis) belongs to which ML type?

SCORE

0 / 3