Calculation of average precision

[tshu-w]

According to the definition of average precision from wikipedia or sklearn, AP computes the average value of p(r) over the interval from r=0 to r=1, i.e., the area under precision-recall curve.

However, the following code calculates the mean of precision directly, i.e., the area under p-k curve.

starmie/checkPrecisionRecall.py

Lines 90 to 94 in 00cae40

	map_sum = 0
	for k in range(0, max_k):
	map_sum += precision_array[k]
	mean_avg_pr = map_sum/max_k
	print("The mean average precision is:", mean_avg_pr)

I believe that this may not affect the overall trend of the different methods and the contribution of this paper, just a friendly reminder :)

[jw-megagon]

Thank you so much for pointing it out. We directly use the evaluation metrics from the code in another paper which is a baseline method to compare with. We will add more clarification about it and allow users to customize the evaluation metrics when officially releasing this repo later.