Double Moons - Machine Learning Example

Moons Machine Learning Demo on DECTRIS CLOUD

Overview

This demo lets you train a machine learning model on the “two moons” dataset and then evaluate four points of your choice against the model.

It runs in two stages:

train_moons.py

Generates the two-moons dataset.
Trains an SVM (support vector machine) classifier.
Prints accuracy, saves the trained model and metadata.

interactive_moons.py

Loads the trained model.
Classifies four points (A, B, C, D) that you choose.
Prints predictions and probabilities.
Saves a TXT report and a PNG plot with the decision regions and your chosen points.

Both scripts are already packaged in the DECTRIS CLOUD job template. You don’t have to run them manually.

How to Use in DECTRIS CLOUD

Start the Job Template

Select the “Moons Machine Learning Demo” template in DECTRIS CLOUD.

This template automatically runs train_moons.py with reasonable defaults.

Choose Your 4 Points

The template will prompt you for four coordinates:

Point A (--point-a)

Point B (--point-b)

Point C (--point-c)

Point D (--point-d)

Enter each point as x,y (for example: 0.5,0.0 or -0.2,0.8).

Parentheses are not required. Negative values and zeros are valid (e.g. -1.0,0 or 0,0).

Run the Job

Submit the job; DECTRIS CLOUD takes care of training and evaluation.

When it finishes, check the outputs.

Outputs

The job produces:

moons_model.joblib
The trained model + metadata.

0results.txt
A text file listing your points, predicted classes, and probabilities.

0results.png
A plot of the decision regions, training data, and your four points (shown as big red stars with bold labels).

Example

Suppose you choose these four points:

Point A: 0.5,0.0 Point B: 1.0,-0.5 Point C: -0.2,0.8 Point D: 0,0

Your output will contain lines like:

Point A (0.500, 0.000) → class 1 (p0=0.123, p1=0.877) Point B (1.000,-0.500) → class 0 (p0=0.654, p1=0.346) ...

And the plot will show:

The two moons background (colored decision regions).

Your 4 points marked in red stars with bold labels.

Tips

If you leave a point blank, the script falls back to a default.

Negative values (like -0.5,0.2) and zeros (0,0) are fully supported.