Interpretation of the results

[havershlomi]

Hey I have read the MAML paper a few times already, and i have ran the code in this repo over miniImageNet (classification task), I was trying to get the same results in the few shot scenario 1 and 5, in both i couldn't get the same results (as in the paper) and i wonder if maybe i don't know how to read the results?

I used this repo to get the dataset: https://github.com/yaoyao-liu/mini-imagenet-tools

5 Shot case:

Mean validation accuracy/loss, stddev, and confidence intervals
(array([0.20673339, 0.25913334, 0.2700002 , 0.28293344, 0.29920024,
       0.32380033, 0.34480017, 0.36266708, 0.36813366, 0.36946717,
       0.37060046], dtype=float32), 
array([0.07652438, 0.07922475, 0.07856207, 0.08520987, 0.09059452,
       0.09908354, 0.10491098, 0.10670018, 0.10576959, 0.10661323,
       0.10691902], dtype=float32),
 array([0.00612323, 0.0063393 , 0.00628627, 0.00681821, 0.00724907,
       0.00792833, 0.00839463, 0.00853779, 0.00846333, 0.00853084,
       0.0085553 ], dtype=float32))
Traceback (most recent call last):

Can anyone explain to me how to read those results? which one is the one the paper is referring to?

[havershlomi]

@cbfinn Can you assist here?