Skip to main content

Task type

A task type describes what you want the model to produce.

You can select multiple tasks if you want one system to do more than one thing (multi-task learning).

If you are unsure, start by selecting the task that matches your evaluation:

  • If you will evaluate with “accuracy” → classification
  • If you will evaluate with “error” (MAE/RMSE) → regression/forecasting

Options in this tool​

Classification​

Predict discrete categories, labels, or classes. Examples: disease present/absent, land-cover class, spam/ham.

Regression​

Predict continuous numerical values.
Examples: temperature, risk score, concentration, price.

Segmentation​

Classify pixels within an image Examples: tumor segmentation, cloud masks, road extraction.

Detection​

Locate objects and/or predict bounding boxes.
Examples: vehicle detection, lesion detection, wildfire hotspots.

Forecasting​

Predict future states from past.
Examples: time series, demand forecasting, weather variables, system monitoring.

Generation​

Generate novel data.
Examples: image synthesis, text generation, data augmentation.

Control/decision-making​

Choose actions to optimise an objective.
Examples: control systems, adaptive sampling, scheduling, policy learning, robotics.