Next version¶
We plan to include the following changes to future versions:
- Version 3.0.0
Add Neural Architecture Search capabilities, which would allow it to recommend the optimal architecture to be used in a given task.
Add high-level transfer learning functionalities, which will make it simple to to adapt pre-trained models to new environments and tasks.
Add methods to calibrate neural networks so that the output scores correspond more closely to probabilities/confidences.