🧠 Machine Learning
The whole point of the platform — train, evaluate, rank models.
5 concepts
1
Supervised learning
Train on labelled (X, y) pairs to predict y for new X.
›
2
Train/test split
Hold out part of the data to measure generalisation. Never evaluate on training data.
›
3
Evaluation metrics
Accuracy, precision, recall, F1, RMSE, MAE — pick the metric that matches the goal.
›
4
Pickle / joblib serialisation
Save a trained model to a .pkl file. Load it later without retraining.
›
5
Framework auto-detection
Sniffing the imports in pasted code to suggest the right pip install.
›