Changelog

​

Batches, faster predictions, new methods, and new names

​

Batches

• You can now include batch object when creating multiple predictions to provide a name and description for the set.
• Predictions which are part of a batch will have a batch field set with an ID for the batch.
• This ID can be used to fetch all predictions in the batch: /predictions/batch/{batch_id}

​

Faster predictions

• Predictions are now ~3x faster, taking less than 15 seconds. Watch the demo video (has sound).

​

New methods

• GET /populations returns all population models available for simulation
• GET /predictions|tests/batch/{batch_id} returns batch details and all batch predictions for a batch

​

New names

• POST /answers -> POST /predictions
• POST /answers/benchmark -> POST /tests

• simulated_answer_percentages -> predicted_answer_percentages

• simulated_answer_percentages -> predicted_answer_percentages
• kullback_leibler_divergence -> information_loss
• normalised_kullback_leibler_divergence -> normalised_information_loss
• root_mean_squared_error -> squared_error

• avg_mean_absolute_error -> average_accuracy (inverted)
• avg_mean_squared_error -> average_squared_error
• avg_normalised_kullback_leibler_divergence -> average_normalised_information_loss

​

The Semilattice API: user insights as infrastructure

• Create population models with custom seed data
• Test model accuracy with built-in cross-validation
• Test model accuracy on external ground truth data
• Predict new questions