search

kjunelee / MetaOptNet

Meta-Learning with Differentiable Convex Optimization (CVPR 2019 Oral)
Apache License 2.0
517 stars 95 forks source link

Reproduced results on miniImagenet with prototypical network with 4-layer network #9

Closed kikyou123 closed 5 years ago

kikyou123 commented 5 years ago

Hi, many thanks for the sharing of the code. I reproduced the results of the prototypical network (4-layer network)on miniImagenet, under python=2.7 with following packages. torch==1.0.0.

I trained with only train dataset, and the best model is selected on the validation set. I tested the best model on the test data, the results are: Accuracy: (1-shot) 43.79 ± 0.65 --- your paper: ( 53.47 )

I am wondering if the difference is reasonable in this task? I am also wondering if the difference is due to my running environments.

Many thanks!!

kjunelee commented 5 years ago

I guess it is because you meta-trained with 15-shot or 10-shot episodes. As shown in Figure 2 of our paper, 1-shot performance of ProtoNet drops significantly if you meta-train with high shot.

kikyou123 commented 5 years ago

Yes, I will try with 1-shot. Thanks very much!