Machine Learning NoteBook 0816

8/6: done

9:00 - 12:00 finish the figure of inner test start to do outer test but preform not well

1:00 - 4:00 test the PDB Binding database and find that the accuracy is still increasing and start to train the model again

4:00 - 4:40 revise the inner test figures and data + start the training of model with changing lr

4:40 - 6:00 working on the result part

8:00 - 11:00 train the model, and find a problem in the result, solving

current status

finish the model again
do the inner test with a better precise way
do the outer test with PDB binding database

need to do tomorrow

solve the problem:

notice

you can use scheduler to help optimizer have a better performance, but it needs to change a lot of things

optimizer = optim.Adam(net.parameters(),lr=0.01)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min', factor=0.8, minlr=0.000000001)

# don't forget to update lr
scheduler.step(testmse)

you can check the change of lr with:

1	current_lr = optimizer.param_groups[0]['lr']

figures and reference for paper

figure: the performance of inner test(fig.1)
table: the detail value of the inner test(table.1)

Test Set	$R$	$RMSE$	$MAE$	$R^2$	$MSE$
Inner Test – 149	0.81	0.87	0.65	0.66	0.75
Inner Test – 199	0.83	0.83	0.62	0.69	0.69
Inner Test – 249	0.84	0.82	0.61	0.70	0.67
Inner Test – 289	0.85	0.80	0.59	0.72	0.64

# Code # Notebook # MachineLearning