Feat/mdm predict distances #248
Conversation
- override only the instance method and not the class function
…ull_cpm" add "weights_euclid_cpm"
| @@ -472,7 +472,7 @@ def __init__( | |||
|
|
|||
| def _create_pipe(self): | |||
| clf_mean_logeuclid_dist_cpm = QuantumMDMWithRiemannianPipeline( | |||
There was a problem hiding this comment.
This VotingClassifier should be replaced by a voting classifier Q-MDM/NHC or Q-MDM/MDM
pyriemann_qiskit/utils/distance.py
Outdated
| objectif = prob.sum( | ||
| w[i] * distance(B, A[i]) for i in matrices | ||
| ) | ||
|
|
||
| prob.set_objective("min", objectif) | ||
| prob.add_constraint(prob.sum(w) == 1) | ||
| weights = optimizer.solve(prob, reshape=False) |
There was a problem hiding this comment.
This is pretty basic. We can think of something clever in the future. Let's just put the structure in place.
|
Can you please put a more detailed description of the PR. |
pyriemann_qiskit/classification.py
Outdated
| @@ -16,6 +17,7 @@ | |||
| NaiveQAOAOptimizer, | |||
| set_global_optimizer, | |||
| ) | |||
| from pyriemann_qiskit.utils.distance import distance_functions, is_cpm_dist | |||
There was a problem hiding this comment.
is_cpm_dist should be moved from utils.distance to classification.
There was a problem hiding this comment.
Hm. Yes, but I will need to move is_qmean too, but it is also used in the pipeline module.
There was a problem hiding this comment.
We should merge is_qdist and is_qmean into a generic is_qfunction.
There was a problem hiding this comment.
Ok. Let's create a new module utils/utils.py?
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
complete api.rst
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>
…yQiskitOptimizer`
There was a problem hiding this comment.
Thx @gcattan !
Last remark: don't forget to update whatsnew file.
|
Thx @qbarthelemy :) Until now I was updating the whatsnew on each release only, based on the generated release notes. |
No, don't change anything. |
This PR:
_predict_distancesfrom thedistanceto theclassificationmodule._predict_distancesof the MDM class but only the one of a particular instance.dist_logeuclid_cpmtodist_logeuclid_to_convex_hull_cpmeuclid_cpmandlogeuclid_cpm(the idea would be rather to re-implement the recursion in the pyRiemann distance module in the CPM way)