-
Notifications
You must be signed in to change notification settings - Fork 110
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Goodness-of-fit diagnostics for Hawkes EM algorithm #507
Goodness-of-fit diagnostics for Hawkes EM algorithm #507
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice
@PhilipDeegan Do you know why the build on Ubuntu is failing? |
consider rebasing #510, with that we can see if all the builds fail the same wya |
ulong uv = u * n_nodes + v; | ||
for (ulong k = 0; k < kernel_size; ++k) { | ||
ulong vk = v * kernel_size + k; | ||
double t = t0 + k * dt + .5 * dt; // kernel_discretization[k] + .5 * dt; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
are these helpful in any way?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think so. The evaluation of the Hawkes kernel is very consequential for the EM algorithm. Before these tests were in place, I could not really be sure about the correctness of the evaluation.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sorry I meant the commented out code, not the test
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ah, sure, that was meant to link the manual increment to the point in the array. I can take that out.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done here: e43af7e
This PR proposes to add new methods to the class
HawkesEM
, and to expand goodness-of-fit diagnostics for the expectation-maximisation algorithm. The key capability is the computation of compensators of Hawkes processes, defined as time-integrals of the conditional intensity functions. (Notice that "compensator" is also used to refer to a subroutine in the calibration routine, but no confusion should arise).This PR also introduces google tests for the class
HawkesEM
.