Convonlutional neural network: iterating trough parameters without result

I am currently working on a project in which I use Matlab to perform transferlearning on Alexnet. My dataset is a set of 1500 surgical images, labeled as useful or not useful.

I tried iterating trough four different hyperparameters to obtain the best results. However, the plots (among others: validation accuracy against the hyperparameter of choice) I obtain tend to shoot 'random' from zero to the same value every time. This make it seems that the value of the hyperparameters is not influencing my performance at all. Below is an example of a plot I obtained.

I hope one of you knows what is happening here.

Answer 1

[Myrthe B.]

This is part of my code:

for hyperparameter=1:max_hyperparameter
  options=trainingOptions(hyperparameter) [newnet,info]=trainNetwork 
  pred=classify(newnet,imds_test) accuracy_dev(end_1)=info.ValidationAccuracy(end) 
  accuracy_test(end+1)=sum(pred_test==labels_test)/length(pred_test)*100 end
end