Multimode error analysis! #44
Conversation
|
|
||
| def run( | ||
| self, method: str, *args, parse_output=lambda X: X, num_samples=11, **kwargs | ||
| ): |
There was a problem hiding this comment.
Might need a docstring here to clarify these arguments.
quantumnetworks/analysis/error.py
Outdated
| scalings = np.random.normal(size=num_samples * num_variables).reshape( | ||
| num_samples, num_variables | ||
| ) |
There was a problem hiding this comment.
Minor, but you might be able to specify the shape directly in size, without needing np.reshape.
E.g. size=(num_samples, num_variables)
There was a problem hiding this comment.
Just to confirm, but this is assuming the error on all omegas or kappas, etc. is the same? That is omegas_error is assumed to be the same for omega_a, omega_b, etc? Might be good to mention that explicitly in a docstring since in principle one might have different uncertainties on those (e.g. if omega_a is a microwave resonator and omega_b is a Fabry-Perot optical cavity, the experiments one would do to measure the frequencies are different and so might have different uncertainties).
That said, keeping omegas_error the same for all is totally fine here (i.e. not worth the time to have to explicitly set all uncertainties manually). So maybe just mention it as a specific choice and leave it at that
There was a problem hiding this comment.
No, omegas_error is a vector and we're doing element-wise multiplication when we do omegas_error * scaling_vec[:num_modes]!
Check the notebook for detailed use!
| def calculate_error( | ||
| self, method: str, *args, num_samples=11, parse_output=lambda X: X, **kwargs | ||
| ): |
There was a problem hiding this comment.
Copying the description over from #4, but good to add a docstring here to clarify the general idea of what this method is trying to do.
General idea:
Let's say we measure parametera ± δa,b ± δb, andc ± δc. Then, to find the error inf(a,b,c), we can sample the parameter values ofa*in the gaussian distribution centered aroundawith standard deviation ofδa(and similarly forb*andc*) and calculatef(a*,b*,c*)multiple times. Then, we can take the standard deviation of that set off(a*,b*,c*)s to findδf.
|
Comments mainly consist of adding docstrings and other minor fixes. Otherwise, looks solid! 💯 |
PR to close #4 .
Relevant notebook: demos/analysis/ErrorAnalysis.ipynb