Learner Utils

This section contains helper code for the Octopod learner pipelines for supporting multiple loss functions and metrics for individual tasks.

Metric Utils

octopod.learner_utils.metrics_utils.multi_class_accuracy(y_true, y_raw_preds)

Takes in raw outputs from Octopod task heads and outputs an accuracy metric and the processed predictions after a softmax as been applied

Parameters
  • y_true (np.array) – Target labels for a specific task for the predicted samples in y_raw_preds

  • y_raw_preds (np.array) – predicted values for the validation set for a specific task

Returns

  • acc (float) – Output of a sklearn accuracy score function

  • y_preds (np.array) – array of predicted values where a softmax has been applied

octopod.learner_utils.metrics_utils.multi_label_accuracy(y_true, y_raw_preds)

Takes in raw outputs from Octopod task heads and outputs an accuracy metric and the processed predictions after a sigmoid as been applied

Parameters
  • y_true (np.array) – Target labels for a specific task for the predicted samples in y_raw_preds

  • y_raw_preds (np.array) – predicted values for the validation set for a specific task

Returns

  • acc (float) – Output of a sklearn accuracy score function

  • y_preds (np.array) – array of predicted values where a sigmoid has been applied