Acquasition fun

[rishabh-mondal]

Implementing Maximum Entropy Acquisition for Ensemble Methods and Monte Carlo Dropout Sampling Strategies and test cases

[nipunbatra]

Might it be better to use: https://pytorch.org/docs/stable/generated/torch.nn.LogSoftmax.html

cc @patel-zeel

[nipunbatra]

Acquisition is spelled wrong in the PR title. Moreover, I don't know if fun stands for function or something else.

[coveralls]

Pull Request Test Coverage Report for Build 6678753082

• 0 of 8 (0.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-1.1%) to 59.17%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
astra/torch/al/acquisitions/max_entropy.py	0	8	0.0%

Totals
Change from base Build 6676378766:	-1.1%
Covered Lines:	271
Relevant Lines:	458

---

💛 - Coveralls

[patel-zeel]

@rishabh-mondal A few initial comments:

• As I mentioned in Create acquisition functions #3, shape needs to be checked differently for Ensemble/MC and Deterministic. So, this function should also inherit DeterministicAcquisition.
• Implement this function with https://pytorch.org/docs/stable/generated/torch.nn.LogSoftmax.html as suggested above.
• Should we call it MaxEntropy or just Entropy? I'd go with the name Entropy. By default, all acquisitions are maximized.
• Tests should be as a .py file, and it should tested with pytest locally.