Implement entropy change measure

[sachinn12]

Implement a measure of change like apd_compare_all as entropy of the density of wic predictions.

Generally: Copy APD function, get histogram of similarities, get entropy of histogram.

More specifically: implement these different measures:

• H_COMPARE: H(COMPARE)
• H_ALL: H(ALL)
• H_ALL_COMPARE: H(ALL)-H(COMPARE)
• H_COMPARE_EARLIER_LATER: H(COMPARE)-0.5*((H(EARLIER)+H(LATER)))
• H_ALL_EARLIER_LATER: H(ALL)-0.5*((H(EARLIER)+H(LATER)))

where H is entropy of the histogram of similarities. Keep the number of bins fixed to some number, maybe bins=10 or 20, we need to look at some histograms to decide that, i.e., please plot some of these histograms and paste them here in the issue.

Please test the above-described measures on the English and German DWUG datasets and report graded change performance here.

After implementation discuss the code.