Expectation of prediction time

[SZH1230456]

Hi, I am confused about the expectation calculation of prediction time. In the paper, the authors just mentioned using the numerical integration technique method. Before feeding the time into the model, what operations did you carry out? Maybe using min-max normalization or something else? I used the code you providing but obtained a very large error in terms of MAE or MSE. Could you please explain it? Thanks.

[musically-ut]

The predict method at https://github.com/musically-ut/tf_rmtpp/blob/master/src/tf_rmtpp/rmtpp_core.py#L575 does the quad integration. This follows the same method as in the paper.

[SZH1230456]

Thank you. I have seen it. But what really confused me a lot lies in the data and result. I directly fed real-time with no Min-Max or other transformations and got a large MSE and MAE error. Could you please tell me how did you process the time before feeding into the model?

[musically-ut]

I see. TBH, I do not exactly remember if any particular normalization was needed outside of what the code already does.

You could have a look at the original implementation https://github.com/dunan/NeuralPointProcess for some hints too (IIRC, there are minor differences in the exact implementation).

[SZH1230456]

Ok. Thanks a lot.