Irreproducible Example_1_function_fitting.ipynb

[tyagi-ankit]

Example_1_function_fitting.ipynb runs, but it's plots/results no longer match the saved ones on github. They also don't match the ones given in the paper.

Results obtained by running Example_1_function_fitting.ipynb are very poor.

[junfeng-chen]

Example_1_function_fitting.ipynb runs, but it's plots/results no longer match the saved ones on github. They also don't match the ones given in the paper.

Results obtained by running Example_1_function_fitting.ipynb are very poor.

I had similar results to yours. After changing the width to [2,4,1] and fitting the model on GPU, I managed to obtain a loss at the level of 1e-5.

[tyagi-ankit]

I found the results change drastically if you change the seed, while keeping everything else the same.