DARTsort 0.3.5__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- dartsort/__init__.py +68 -0
- dartsort/cli.py +105 -0
- dartsort/clustering/__init__.py +3 -0
- dartsort/clustering/cluster_util.py +392 -0
- dartsort/clustering/clustering.py +765 -0
- dartsort/clustering/clustering_features.py +202 -0
- dartsort/clustering/density.py +1299 -0
- dartsort/clustering/forward_backward.py +219 -0
- dartsort/clustering/gmm/mixture.py +5790 -0
- dartsort/clustering/gmm/stable_features.py +993 -0
- dartsort/clustering/kmeans.py +448 -0
- dartsort/clustering/merge.py +501 -0
- dartsort/clustering/modes.py +214 -0
- dartsort/clustering/refine_util.py +191 -0
- dartsort/clustering/relocate.py +109 -0
- dartsort/clustering/split.py +1285 -0
- dartsort/config.py +288 -0
- dartsort/detect/__init__.py +2 -0
- dartsort/detect/detect.py +449 -0
- dartsort/detect/detect_filters.py +37 -0
- dartsort/evaluate/__init__.py +1 -0
- dartsort/evaluate/analysis.py +594 -0
- dartsort/evaluate/comparison.py +534 -0
- dartsort/evaluate/hybrid_util.py +599 -0
- dartsort/evaluate/noise_recording_tools.py +184 -0
- dartsort/evaluate/sim_template_tools.py +842 -0
- dartsort/evaluate/simkit.py +840 -0
- dartsort/evaluate/simlib.py +270 -0
- dartsort/localize/__init__.py +0 -0
- dartsort/localize/localize_torch.py +291 -0
- dartsort/localize/localize_util.py +313 -0
- dartsort/main.py +657 -0
- dartsort/peel/__init__.py +6 -0
- dartsort/peel/grab.py +208 -0
- dartsort/peel/matching.py +628 -0
- dartsort/peel/matching_util/__init__.py +3 -0
- dartsort/peel/matching_util/compressed_upsampled.py +718 -0
- dartsort/peel/matching_util/drifty.py +791 -0
- dartsort/peel/matching_util/matching_base.py +616 -0
- dartsort/peel/matching_util/pairwise.py +359 -0
- dartsort/peel/matching_util/pairwise_util.py +1317 -0
- dartsort/peel/peel_base.py +1000 -0
- dartsort/peel/running_template.py +1498 -0
- dartsort/peel/subtract.py +1123 -0
- dartsort/peel/threshold.py +471 -0
- dartsort/peel/universal.py +191 -0
- dartsort/pretrained/default_temporal_kernel.npy +0 -0
- dartsort/pretrained/detect_np1.pt +0 -0
- dartsort/pretrained/detect_np2.pt +0 -0
- dartsort/pretrained/glasbey1024.npz +0 -0
- dartsort/pretrained/offset_trained_denoiser.pt +0 -0
- dartsort/pretrained/offset_trained_denoiser_added_temp.pt +0 -0
- dartsort/pretrained/offset_trained_denoiser_added_temp_boost.pt +0 -0
- dartsort/pretrained/original_single_chan_denoiser.pt +0 -0
- dartsort/pretrained/single_chan_denoiser.pt +0 -0
- dartsort/templates/__init__.py +11 -0
- dartsort/templates/get_templates.py +771 -0
- dartsort/templates/postprocess_util.py +555 -0
- dartsort/templates/realignment.py +466 -0
- dartsort/templates/superres_util.py +204 -0
- dartsort/templates/template_util.py +518 -0
- dartsort/templates/templates.py +528 -0
- dartsort/transform/__init__.py +2 -0
- dartsort/transform/_multichan_denoiser_kit.py +650 -0
- dartsort/transform/all_transformers.py +58 -0
- dartsort/transform/amortized_localization.py +503 -0
- dartsort/transform/amplitudes.py +197 -0
- dartsort/transform/decollider.py +635 -0
- dartsort/transform/enforce_decrease.py +158 -0
- dartsort/transform/localize.py +86 -0
- dartsort/transform/matching_denoiser.py +44 -0
- dartsort/transform/pipeline.py +391 -0
- dartsort/transform/single_channel_denoiser.py +153 -0
- dartsort/transform/single_channel_templates.py +98 -0
- dartsort/transform/supervised_denoiser.py +486 -0
- dartsort/transform/temporal_pca.py +413 -0
- dartsort/transform/transform_base.py +255 -0
- dartsort/util/__init__.py +0 -0
- dartsort/util/cli_util.py +179 -0
- dartsort/util/data_util.py +1128 -0
- dartsort/util/drift_util.py +1054 -0
- dartsort/util/internal_config.py +1074 -0
- dartsort/util/interpolation_util.py +1012 -0
- dartsort/util/job_util.py +19 -0
- dartsort/util/logging_util.py +91 -0
- dartsort/util/main_util.py +235 -0
- dartsort/util/more_operators.py +269 -0
- dartsort/util/multiprocessing_util.py +211 -0
- dartsort/util/nn_util.py +461 -0
- dartsort/util/noise_util.py +1461 -0
- dartsort/util/peel_util.py +223 -0
- dartsort/util/py_util.py +213 -0
- dartsort/util/registration_util.py +128 -0
- dartsort/util/residual_util.py +30 -0
- dartsort/util/sparse_util.py +81 -0
- dartsort/util/spikeio.py +227 -0
- dartsort/util/spiketorch.py +1198 -0
- dartsort/util/testing_util/__init__.py +0 -0
- dartsort/util/testing_util/matching_debug_util.py +517 -0
- dartsort/util/testing_util/mixture_testing_util.py +291 -0
- dartsort/util/torch_optimization_util.py +156 -0
- dartsort/util/torch_util.py +75 -0
- dartsort/util/universal_util.py +269 -0
- dartsort/util/waveform_util.py +809 -0
- dartsort/vis/__init__.py +19 -0
- dartsort/vis/analysis_plots.py +392 -0
- dartsort/vis/colors.py +36 -0
- dartsort/vis/gt.py +392 -0
- dartsort/vis/layout.py +149 -0
- dartsort/vis/metrics_util.py +23 -0
- dartsort/vis/mixture.py +1489 -0
- dartsort/vis/over_time.py +514 -0
- dartsort/vis/recanim.py +429 -0
- dartsort/vis/scatterplots.py +779 -0
- dartsort/vis/sorting.py +163 -0
- dartsort/vis/unit.py +917 -0
- dartsort/vis/unit_comparison.py +1017 -0
- dartsort/vis/versus.py +300 -0
- dartsort/vis/vismain.py +537 -0
- dartsort/vis/waveforms.py +357 -0
- dartsort-0.3.5.dist-info/METADATA +93 -0
- dartsort-0.3.5.dist-info/RECORD +125 -0
- dartsort-0.3.5.dist-info/WHEEL +5 -0
- dartsort-0.3.5.dist-info/entry_points.txt +2 -0
- dartsort-0.3.5.dist-info/top_level.txt +1 -0
|
@@ -0,0 +1,765 @@
|
|
|
1
|
+
import gc
|
|
2
|
+
from typing import cast, Literal, Self
|
|
3
|
+
|
|
4
|
+
import numpy as np
|
|
5
|
+
import sklearn.cluster
|
|
6
|
+
import torch
|
|
7
|
+
from spikeinterface.core import BaseRecording
|
|
8
|
+
|
|
9
|
+
from ..util.data_util import chunk_time_ranges, fit_reweighting, DARTsortSorting
|
|
10
|
+
from ..util.main_util import ds_save_intermediate_labels
|
|
11
|
+
from ..util.internal_config import (
|
|
12
|
+
ClusteringConfig,
|
|
13
|
+
RefinementConfig,
|
|
14
|
+
FitSamplingConfig,
|
|
15
|
+
ComputationConfig,
|
|
16
|
+
)
|
|
17
|
+
from .clustering_features import SimpleMatrixFeatures
|
|
18
|
+
from ..util import job_util
|
|
19
|
+
from . import cluster_util, density, forward_backward, refine_util
|
|
20
|
+
from .gmm import mixture
|
|
21
|
+
|
|
22
|
+
|
|
23
|
+
clustering_strategies: dict[str, "type[Clusterer]"] = {}
|
|
24
|
+
refinement_strategies: dict[str, "type[Refinement]"] = {}
|
|
25
|
+
|
|
26
|
+
|
|
27
|
+
def get_clusterer(
|
|
28
|
+
clustering_cfg: ClusteringConfig | None = None,
|
|
29
|
+
refinement_cfgs: list[RefinementConfig | None] | None = None,
|
|
30
|
+
computation_cfg: ComputationConfig | None = None,
|
|
31
|
+
save_cfg=None,
|
|
32
|
+
save_labels_dir=None,
|
|
33
|
+
initial_name=None,
|
|
34
|
+
refine_labels_fmt=None,
|
|
35
|
+
):
|
|
36
|
+
if clustering_cfg is not None:
|
|
37
|
+
if clustering_cfg.cluster_strategy not in clustering_strategies:
|
|
38
|
+
raise ValueError(
|
|
39
|
+
f"Unknown cluster_strategy={clustering_cfg.cluster_strategy}. "
|
|
40
|
+
f"Options are: {', '.join(clustering_strategies.keys())}."
|
|
41
|
+
)
|
|
42
|
+
clus_strategy = clustering_cfg.cluster_strategy
|
|
43
|
+
else:
|
|
44
|
+
clus_strategy = "none"
|
|
45
|
+
|
|
46
|
+
saving_labels = save_cfg is not None and save_cfg.save_intermediate_labels
|
|
47
|
+
if saving_labels:
|
|
48
|
+
shared_save_kw = dict(
|
|
49
|
+
save_labels_dir=save_labels_dir,
|
|
50
|
+
save_cfg=save_cfg,
|
|
51
|
+
)
|
|
52
|
+
else:
|
|
53
|
+
shared_save_kw = dict(
|
|
54
|
+
save_labels_dir=None,
|
|
55
|
+
save_cfg=None,
|
|
56
|
+
)
|
|
57
|
+
|
|
58
|
+
C = clustering_strategies[clus_strategy]
|
|
59
|
+
init_fmt = initial_name if (saving_labels and clustering_cfg is not None) else None
|
|
60
|
+
clusterer = C.from_config(
|
|
61
|
+
clustering_cfg,
|
|
62
|
+
labels_fmt=init_fmt,
|
|
63
|
+
computation_cfg=computation_cfg,
|
|
64
|
+
**shared_save_kw,
|
|
65
|
+
)
|
|
66
|
+
|
|
67
|
+
refinement_cfgs = refinement_cfgs or []
|
|
68
|
+
for j, r_cfg in enumerate(refinement_cfgs):
|
|
69
|
+
if r_cfg is None:
|
|
70
|
+
continue
|
|
71
|
+
|
|
72
|
+
strat = r_cfg.refinement_strategy
|
|
73
|
+
if strat not in refinement_strategies:
|
|
74
|
+
raise ValueError(
|
|
75
|
+
f"Unknown refinement_strategy={strat}. "
|
|
76
|
+
f"Options are: {', '.join(refinement_strategies.keys())}."
|
|
77
|
+
)
|
|
78
|
+
|
|
79
|
+
R = refinement_strategies[strat]
|
|
80
|
+
rsave = saving_labels and strat != "none"
|
|
81
|
+
if rsave and refine_labels_fmt is not None:
|
|
82
|
+
stepfmt = f"_{j}_{strat}_{{stepname}}"
|
|
83
|
+
r_fmt = refine_labels_fmt.format(stepname=stepfmt)
|
|
84
|
+
else:
|
|
85
|
+
r_fmt = None
|
|
86
|
+
|
|
87
|
+
clusterer = R(
|
|
88
|
+
clusterer,
|
|
89
|
+
refinement_cfg=r_cfg,
|
|
90
|
+
labels_fmt=r_fmt,
|
|
91
|
+
computation_cfg=computation_cfg,
|
|
92
|
+
**shared_save_kw,
|
|
93
|
+
)
|
|
94
|
+
|
|
95
|
+
return clusterer
|
|
96
|
+
|
|
97
|
+
|
|
98
|
+
class Clusterer:
|
|
99
|
+
def __init__(
|
|
100
|
+
self,
|
|
101
|
+
computation_cfg: ComputationConfig | None = None,
|
|
102
|
+
sampling_cfg: FitSamplingConfig | None = None,
|
|
103
|
+
save_cfg=None,
|
|
104
|
+
save_labels_dir=None,
|
|
105
|
+
labels_fmt=None,
|
|
106
|
+
):
|
|
107
|
+
self.computation_cfg = computation_cfg
|
|
108
|
+
self.sampling_cfg = sampling_cfg
|
|
109
|
+
if computation_cfg is None:
|
|
110
|
+
self.computation_cfg = job_util.get_global_computation_config()
|
|
111
|
+
self.save_cfg = save_cfg
|
|
112
|
+
self.save_labels_dir = save_labels_dir
|
|
113
|
+
self.labels_fmt = labels_fmt
|
|
114
|
+
|
|
115
|
+
@classmethod
|
|
116
|
+
def from_config(
|
|
117
|
+
cls,
|
|
118
|
+
clustering_cfg: ClusteringConfig | None,
|
|
119
|
+
computation_cfg: ComputationConfig | None = None,
|
|
120
|
+
save_cfg=None,
|
|
121
|
+
save_labels_dir=None,
|
|
122
|
+
labels_fmt=None,
|
|
123
|
+
) -> Self:
|
|
124
|
+
if clustering_cfg is None:
|
|
125
|
+
sampling_cfg = None
|
|
126
|
+
else:
|
|
127
|
+
sampling_cfg = clustering_cfg.sampling_cfg
|
|
128
|
+
return cls(
|
|
129
|
+
computation_cfg=computation_cfg,
|
|
130
|
+
save_cfg=save_cfg,
|
|
131
|
+
save_labels_dir=save_labels_dir,
|
|
132
|
+
labels_fmt=labels_fmt,
|
|
133
|
+
sampling_cfg=sampling_cfg,
|
|
134
|
+
)
|
|
135
|
+
|
|
136
|
+
def handle_sampling(
|
|
137
|
+
self, features: SimpleMatrixFeatures
|
|
138
|
+
) -> tuple[Literal[False], slice] | tuple[Literal[True], np.ndarray]:
|
|
139
|
+
if self.sampling_cfg is None:
|
|
140
|
+
return False, slice(None)
|
|
141
|
+
elif features.n <= self.sampling_cfg.n_waveforms_fit:
|
|
142
|
+
return False, slice(None)
|
|
143
|
+
|
|
144
|
+
weights = fit_reweighting(
|
|
145
|
+
voltages=features.signed_amplitudes,
|
|
146
|
+
fit_sampling=self.sampling_cfg.fit_sampling,
|
|
147
|
+
fit_max_reweighting=self.sampling_cfg.fit_max_reweighting,
|
|
148
|
+
)
|
|
149
|
+
rg = np.random.default_rng(self.sampling_cfg.fit_subsampling_random_state)
|
|
150
|
+
ixs = rg.choice(
|
|
151
|
+
features.n,
|
|
152
|
+
size=self.sampling_cfg.n_waveforms_fit,
|
|
153
|
+
p=weights,
|
|
154
|
+
replace=False,
|
|
155
|
+
)
|
|
156
|
+
ixs.sort()
|
|
157
|
+
return True, ixs
|
|
158
|
+
|
|
159
|
+
def cluster(
|
|
160
|
+
self,
|
|
161
|
+
features: SimpleMatrixFeatures,
|
|
162
|
+
sorting: DARTsortSorting,
|
|
163
|
+
recording: BaseRecording,
|
|
164
|
+
motion_est=None,
|
|
165
|
+
) -> DARTsortSorting:
|
|
166
|
+
if features is None:
|
|
167
|
+
pass
|
|
168
|
+
else:
|
|
169
|
+
labels = self._cluster(features, sorting, recording, motion_est)
|
|
170
|
+
sorting = sorting.ephemeral_replace(labels=labels)
|
|
171
|
+
if self.labels_fmt and self.save_labels_dir is not None:
|
|
172
|
+
assert "{" not in self.labels_fmt
|
|
173
|
+
assert "}" not in self.labels_fmt
|
|
174
|
+
ds_save_intermediate_labels(
|
|
175
|
+
self.labels_fmt, sorting, self.save_labels_dir, self.save_cfg
|
|
176
|
+
)
|
|
177
|
+
return sorting
|
|
178
|
+
|
|
179
|
+
def _cluster(
|
|
180
|
+
self,
|
|
181
|
+
features: SimpleMatrixFeatures,
|
|
182
|
+
sorting: DARTsortSorting,
|
|
183
|
+
recording: BaseRecording,
|
|
184
|
+
motion_est=None,
|
|
185
|
+
) -> np.ndarray:
|
|
186
|
+
"""Unused method but shows API."""
|
|
187
|
+
del features, recording, motion_est
|
|
188
|
+
assert sorting.labels is not None
|
|
189
|
+
return sorting.labels
|
|
190
|
+
|
|
191
|
+
|
|
192
|
+
clustering_strategies["none"] = Clusterer
|
|
193
|
+
|
|
194
|
+
|
|
195
|
+
class ChannelSnapClusterer(Clusterer):
|
|
196
|
+
def _cluster(
|
|
197
|
+
self,
|
|
198
|
+
features: SimpleMatrixFeatures,
|
|
199
|
+
sorting: DARTsortSorting,
|
|
200
|
+
recording: BaseRecording,
|
|
201
|
+
motion_est=None,
|
|
202
|
+
) -> np.ndarray:
|
|
203
|
+
return cluster_util.closest_registered_channels(
|
|
204
|
+
times_seconds=sorting.times_seconds, # type: ignore
|
|
205
|
+
x=features.x,
|
|
206
|
+
z_abs=features.z,
|
|
207
|
+
z_reg=features.z_reg,
|
|
208
|
+
geom=recording.get_channel_locations(),
|
|
209
|
+
motion_est=motion_est,
|
|
210
|
+
)
|
|
211
|
+
|
|
212
|
+
|
|
213
|
+
clustering_strategies["channel_snap"] = ChannelSnapClusterer
|
|
214
|
+
|
|
215
|
+
|
|
216
|
+
class GridSnapClusterer(Clusterer):
|
|
217
|
+
def __init__(self, grid_dx=15.0, grid_dz=15.0, **kwargs):
|
|
218
|
+
super().__init__(**kwargs)
|
|
219
|
+
self.grid_dx = grid_dx
|
|
220
|
+
self.grid_dz = grid_dz
|
|
221
|
+
|
|
222
|
+
@classmethod
|
|
223
|
+
def from_config(
|
|
224
|
+
cls,
|
|
225
|
+
clustering_cfg: ClusteringConfig | None,
|
|
226
|
+
computation_cfg: ComputationConfig | None = None,
|
|
227
|
+
save_cfg=None,
|
|
228
|
+
save_labels_dir=None,
|
|
229
|
+
labels_fmt=None,
|
|
230
|
+
) -> Self:
|
|
231
|
+
assert clustering_cfg is not None
|
|
232
|
+
return cls(
|
|
233
|
+
grid_dx=clustering_cfg.grid_dx,
|
|
234
|
+
grid_dz=clustering_cfg.grid_dz,
|
|
235
|
+
computation_cfg=computation_cfg,
|
|
236
|
+
save_cfg=save_cfg,
|
|
237
|
+
save_labels_dir=save_labels_dir,
|
|
238
|
+
labels_fmt=labels_fmt,
|
|
239
|
+
sampling_cfg=clustering_cfg.sampling_cfg,
|
|
240
|
+
)
|
|
241
|
+
|
|
242
|
+
def _cluster(
|
|
243
|
+
self,
|
|
244
|
+
features: SimpleMatrixFeatures,
|
|
245
|
+
sorting: DARTsortSorting,
|
|
246
|
+
recording: BaseRecording,
|
|
247
|
+
motion_est=None,
|
|
248
|
+
) -> np.ndarray:
|
|
249
|
+
return cluster_util.grid_snap(
|
|
250
|
+
times_seconds=sorting.times_seconds, # type: ignore
|
|
251
|
+
x=features.x,
|
|
252
|
+
z_abs=features.z,
|
|
253
|
+
z_reg=features.z_reg,
|
|
254
|
+
grid_dx=self.grid_dx,
|
|
255
|
+
grid_dz=self.grid_dz,
|
|
256
|
+
geom=recording.get_channel_locations(),
|
|
257
|
+
motion_est=motion_est,
|
|
258
|
+
)
|
|
259
|
+
|
|
260
|
+
|
|
261
|
+
clustering_strategies["grid_snap"] = GridSnapClusterer
|
|
262
|
+
|
|
263
|
+
|
|
264
|
+
class DensityPeaksClusterer(Clusterer):
|
|
265
|
+
def __init__(
|
|
266
|
+
self,
|
|
267
|
+
knn_k=None,
|
|
268
|
+
sigma_local=5.0,
|
|
269
|
+
sigma_regional: float | None = 25.0,
|
|
270
|
+
n_neighbors_search=20,
|
|
271
|
+
radius_search=25.0,
|
|
272
|
+
remove_clusters_smaller_than=50,
|
|
273
|
+
noise_density=0.0,
|
|
274
|
+
outlier_radius=5.0,
|
|
275
|
+
outlier_neighbor_count=5,
|
|
276
|
+
workers=-1,
|
|
277
|
+
uhdversion=False,
|
|
278
|
+
random_seed=0,
|
|
279
|
+
**kwargs,
|
|
280
|
+
):
|
|
281
|
+
super().__init__(**kwargs)
|
|
282
|
+
self.knn_k = knn_k
|
|
283
|
+
self.uhdversion = uhdversion
|
|
284
|
+
self.sigma_local = sigma_local
|
|
285
|
+
self.sigma_regional = sigma_regional
|
|
286
|
+
self.n_neighbors_search = n_neighbors_search
|
|
287
|
+
self.radius_search = radius_search
|
|
288
|
+
self.remove_clusters_smaller_than = remove_clusters_smaller_than
|
|
289
|
+
self.noise_density = noise_density
|
|
290
|
+
self.outlier_radius = outlier_radius
|
|
291
|
+
self.outlier_neighbor_count = outlier_neighbor_count
|
|
292
|
+
self.workers = workers
|
|
293
|
+
self.uhdversion = uhdversion
|
|
294
|
+
self.random_seed = random_seed
|
|
295
|
+
|
|
296
|
+
@classmethod
|
|
297
|
+
def from_config(
|
|
298
|
+
cls,
|
|
299
|
+
clustering_cfg: ClusteringConfig | None,
|
|
300
|
+
computation_cfg: ComputationConfig | None = None,
|
|
301
|
+
save_cfg=None,
|
|
302
|
+
save_labels_dir=None,
|
|
303
|
+
labels_fmt=None,
|
|
304
|
+
) -> Self:
|
|
305
|
+
assert clustering_cfg is not None
|
|
306
|
+
uhdversion = clustering_cfg.cluster_strategy == "density_peaks_uhdversion"
|
|
307
|
+
return cls(
|
|
308
|
+
knn_k=clustering_cfg.knn_k,
|
|
309
|
+
sigma_local=clustering_cfg.sigma_local,
|
|
310
|
+
sigma_regional=clustering_cfg.sigma_regional,
|
|
311
|
+
n_neighbors_search=clustering_cfg.n_neighbors_search,
|
|
312
|
+
radius_search=clustering_cfg.radius_search,
|
|
313
|
+
remove_clusters_smaller_than=clustering_cfg.min_cluster_size,
|
|
314
|
+
noise_density=clustering_cfg.noise_density,
|
|
315
|
+
random_seed=clustering_cfg.random_seed,
|
|
316
|
+
outlier_radius=clustering_cfg.outlier_radius,
|
|
317
|
+
outlier_neighbor_count=clustering_cfg.outlier_neighbor_count,
|
|
318
|
+
workers=clustering_cfg.workers,
|
|
319
|
+
uhdversion=uhdversion,
|
|
320
|
+
computation_cfg=computation_cfg,
|
|
321
|
+
save_cfg=save_cfg,
|
|
322
|
+
save_labels_dir=save_labels_dir,
|
|
323
|
+
labels_fmt=labels_fmt,
|
|
324
|
+
sampling_cfg=clustering_cfg.sampling_cfg,
|
|
325
|
+
)
|
|
326
|
+
|
|
327
|
+
def _cluster(
|
|
328
|
+
self,
|
|
329
|
+
features: SimpleMatrixFeatures,
|
|
330
|
+
sorting: DARTsortSorting,
|
|
331
|
+
recording: BaseRecording,
|
|
332
|
+
motion_est=None,
|
|
333
|
+
) -> np.ndarray:
|
|
334
|
+
subsampling, ixs = self.handle_sampling(features)
|
|
335
|
+
X = features.features
|
|
336
|
+
X_fit = X[ixs]
|
|
337
|
+
|
|
338
|
+
if not self.uhdversion:
|
|
339
|
+
res = density.density_peaks(
|
|
340
|
+
X_fit,
|
|
341
|
+
knn_k=self.knn_k,
|
|
342
|
+
sigma_local=self.sigma_local,
|
|
343
|
+
sigma_regional=self.sigma_regional,
|
|
344
|
+
n_neighbors_search=self.n_neighbors_search,
|
|
345
|
+
radius_search=self.radius_search,
|
|
346
|
+
remove_clusters_smaller_than=0,
|
|
347
|
+
noise_density=self.noise_density,
|
|
348
|
+
outlier_radius=self.outlier_radius,
|
|
349
|
+
outlier_neighbor_count=self.outlier_neighbor_count,
|
|
350
|
+
workers=self.workers,
|
|
351
|
+
)
|
|
352
|
+
else:
|
|
353
|
+
res = density.density_peaks_fancy(
|
|
354
|
+
features.xyza,
|
|
355
|
+
features.amplitudes,
|
|
356
|
+
sorting,
|
|
357
|
+
motion_est,
|
|
358
|
+
recording.get_channel_locations(),
|
|
359
|
+
sigma_local=self.sigma_local,
|
|
360
|
+
sigma_regional=self.sigma_regional,
|
|
361
|
+
n_neighbors_search=self.n_neighbors_search,
|
|
362
|
+
radius_search=self.radius_search,
|
|
363
|
+
remove_clusters_smaller_than=self.remove_clusters_smaller_than,
|
|
364
|
+
noise_density=self.noise_density,
|
|
365
|
+
outlier_radius=self.outlier_radius,
|
|
366
|
+
outlier_neighbor_count=self.outlier_neighbor_count,
|
|
367
|
+
workers=self.workers,
|
|
368
|
+
)
|
|
369
|
+
|
|
370
|
+
if subsampling:
|
|
371
|
+
kdtree = res["kdtree"]
|
|
372
|
+
assert isinstance(ixs, np.ndarray)
|
|
373
|
+
rest = np.setdiff1d(np.arange(len(X)), ixs)
|
|
374
|
+
other_labels = density.nearest_neighbor_assign(
|
|
375
|
+
kdtree,
|
|
376
|
+
res["labels"],
|
|
377
|
+
X[rest],
|
|
378
|
+
radius_search=self.radius_search,
|
|
379
|
+
workers=self.workers,
|
|
380
|
+
)
|
|
381
|
+
labels = cluster_util.combine_disjoint(
|
|
382
|
+
ixs, res["labels"], rest, other_labels
|
|
383
|
+
)
|
|
384
|
+
else:
|
|
385
|
+
labels = res["labels"]
|
|
386
|
+
|
|
387
|
+
labels = cluster_util.decrumb(
|
|
388
|
+
labels, min_size=self.remove_clusters_smaller_than, in_place=True
|
|
389
|
+
)
|
|
390
|
+
|
|
391
|
+
return labels
|
|
392
|
+
|
|
393
|
+
|
|
394
|
+
clustering_strategies["dpc"] = DensityPeaksClusterer
|
|
395
|
+
clustering_strategies["density_peaks_uhdversion"] = DensityPeaksClusterer
|
|
396
|
+
|
|
397
|
+
|
|
398
|
+
class GMMDensityPeaksClusterer(Clusterer):
|
|
399
|
+
def __init__(
|
|
400
|
+
self,
|
|
401
|
+
outlier_neighbor_count=10,
|
|
402
|
+
outlier_radius=25.0,
|
|
403
|
+
remove_clusters_smaller_than=50,
|
|
404
|
+
workers=-1,
|
|
405
|
+
n_initializations=5,
|
|
406
|
+
n_iter=50,
|
|
407
|
+
max_components_per_channel=20,
|
|
408
|
+
min_spikes_per_component=10,
|
|
409
|
+
random_state=0,
|
|
410
|
+
use_hellinger=True,
|
|
411
|
+
kmeanspp_min_dist=5.0,
|
|
412
|
+
hellinger_cutoff=0.8,
|
|
413
|
+
hellinger_strong=0.0,
|
|
414
|
+
hellinger_weak=0.0,
|
|
415
|
+
mop=False,
|
|
416
|
+
n_neighbors_search=20,
|
|
417
|
+
max_sigma=5.0,
|
|
418
|
+
max_samples=2_000_000,
|
|
419
|
+
**kwargs,
|
|
420
|
+
):
|
|
421
|
+
super().__init__(**kwargs)
|
|
422
|
+
self.outlier_neighbor_count = outlier_neighbor_count
|
|
423
|
+
self.outlier_radius = outlier_radius
|
|
424
|
+
self.remove_clusters_smaller_than = remove_clusters_smaller_than
|
|
425
|
+
self.workers = workers
|
|
426
|
+
self.n_initializations = n_initializations
|
|
427
|
+
self.n_iter = n_iter
|
|
428
|
+
self.max_components_per_channel = max_components_per_channel
|
|
429
|
+
self.min_spikes_per_component = min_spikes_per_component
|
|
430
|
+
self.random_state = random_state
|
|
431
|
+
self.use_hellinger = use_hellinger
|
|
432
|
+
self.hellinger_cutoff = hellinger_cutoff
|
|
433
|
+
self.hellinger_strong = hellinger_strong
|
|
434
|
+
self.hellinger_weak = hellinger_weak
|
|
435
|
+
self.mop = mop
|
|
436
|
+
self.max_samples = max_samples
|
|
437
|
+
self.n_neighbors_search = n_neighbors_search
|
|
438
|
+
|
|
439
|
+
# unconfigured params
|
|
440
|
+
self.kmeanspp_min_dist = kmeanspp_min_dist
|
|
441
|
+
self.max_sigma = max_sigma
|
|
442
|
+
|
|
443
|
+
@classmethod
|
|
444
|
+
def from_config(
|
|
445
|
+
cls,
|
|
446
|
+
clustering_cfg: ClusteringConfig | None,
|
|
447
|
+
computation_cfg: ComputationConfig | None = None,
|
|
448
|
+
save_cfg=None,
|
|
449
|
+
save_labels_dir=None,
|
|
450
|
+
labels_fmt=None,
|
|
451
|
+
) -> Self:
|
|
452
|
+
assert clustering_cfg is not None
|
|
453
|
+
return cls(
|
|
454
|
+
outlier_neighbor_count=clustering_cfg.outlier_neighbor_count,
|
|
455
|
+
outlier_radius=clustering_cfg.outlier_radius,
|
|
456
|
+
remove_clusters_smaller_than=clustering_cfg.min_cluster_size,
|
|
457
|
+
workers=clustering_cfg.workers,
|
|
458
|
+
n_initializations=clustering_cfg.kmeanspp_initializations,
|
|
459
|
+
n_iter=clustering_cfg.kmeans_iter,
|
|
460
|
+
max_components_per_channel=clustering_cfg.components_per_channel,
|
|
461
|
+
random_state=clustering_cfg.random_seed,
|
|
462
|
+
use_hellinger=clustering_cfg.use_hellinger,
|
|
463
|
+
hellinger_cutoff=clustering_cfg.component_overlap,
|
|
464
|
+
hellinger_strong=clustering_cfg.hellinger_strong,
|
|
465
|
+
hellinger_weak=clustering_cfg.hellinger_weak,
|
|
466
|
+
mop=clustering_cfg.mop,
|
|
467
|
+
max_sigma=clustering_cfg.gmmdpc_max_sigma,
|
|
468
|
+
max_samples=clustering_cfg.sampling_cfg.n_waveforms_fit,
|
|
469
|
+
n_neighbors_search=clustering_cfg.n_neighbors_search,
|
|
470
|
+
computation_cfg=computation_cfg,
|
|
471
|
+
save_cfg=save_cfg,
|
|
472
|
+
save_labels_dir=save_labels_dir,
|
|
473
|
+
labels_fmt=labels_fmt,
|
|
474
|
+
sampling_cfg=clustering_cfg.sampling_cfg,
|
|
475
|
+
)
|
|
476
|
+
|
|
477
|
+
def _cluster(
|
|
478
|
+
self,
|
|
479
|
+
features: SimpleMatrixFeatures,
|
|
480
|
+
sorting: DARTsortSorting,
|
|
481
|
+
recording: BaseRecording,
|
|
482
|
+
motion_est=None,
|
|
483
|
+
) -> np.ndarray:
|
|
484
|
+
res = density.gmm_density_peaks(
|
|
485
|
+
X=features.features,
|
|
486
|
+
channels=sorting.channels,
|
|
487
|
+
outlier_neighbor_count=self.outlier_neighbor_count,
|
|
488
|
+
outlier_radius=self.outlier_radius,
|
|
489
|
+
remove_clusters_smaller_than=self.remove_clusters_smaller_than,
|
|
490
|
+
workers=self.workers,
|
|
491
|
+
n_initializations=self.n_initializations,
|
|
492
|
+
n_iter=self.n_iter,
|
|
493
|
+
max_components_per_channel=self.max_components_per_channel,
|
|
494
|
+
min_spikes_per_component=self.min_spikes_per_component,
|
|
495
|
+
random_state=self.random_state,
|
|
496
|
+
kmeanspp_min_dist=self.kmeanspp_min_dist,
|
|
497
|
+
use_hellinger=self.use_hellinger,
|
|
498
|
+
hellinger_cutoff=self.hellinger_cutoff,
|
|
499
|
+
hellinger_strong=self.hellinger_strong,
|
|
500
|
+
hellinger_weak=self.hellinger_weak,
|
|
501
|
+
n_neighbors_search=self.n_neighbors_search,
|
|
502
|
+
mop=self.mop,
|
|
503
|
+
max_sigma=self.max_sigma,
|
|
504
|
+
max_samples=self.max_samples,
|
|
505
|
+
)
|
|
506
|
+
return res["labels"]
|
|
507
|
+
|
|
508
|
+
|
|
509
|
+
clustering_strategies["gmmdpc"] = GMMDensityPeaksClusterer
|
|
510
|
+
|
|
511
|
+
|
|
512
|
+
class RecursiveHDBSCANClusterer(Clusterer):
|
|
513
|
+
def __init__(
|
|
514
|
+
self,
|
|
515
|
+
min_cluster_size=25,
|
|
516
|
+
min_samples=25,
|
|
517
|
+
cluster_selection_epsilon=1,
|
|
518
|
+
recursive=True,
|
|
519
|
+
**kwargs,
|
|
520
|
+
):
|
|
521
|
+
super().__init__(**kwargs)
|
|
522
|
+
self.min_cluster_size = min_cluster_size
|
|
523
|
+
self.min_samples = min_samples
|
|
524
|
+
self.cluster_selection_epsilon = cluster_selection_epsilon
|
|
525
|
+
self.recursive = recursive
|
|
526
|
+
|
|
527
|
+
@classmethod
|
|
528
|
+
def from_config(
|
|
529
|
+
cls,
|
|
530
|
+
clustering_cfg: ClusteringConfig | None,
|
|
531
|
+
computation_cfg: ComputationConfig | None = None,
|
|
532
|
+
save_cfg=None,
|
|
533
|
+
save_labels_dir=None,
|
|
534
|
+
labels_fmt=None,
|
|
535
|
+
) -> Self:
|
|
536
|
+
assert clustering_cfg is not None
|
|
537
|
+
return cls(
|
|
538
|
+
min_cluster_size=clustering_cfg.min_cluster_size,
|
|
539
|
+
min_samples=clustering_cfg.min_samples,
|
|
540
|
+
cluster_selection_epsilon=clustering_cfg.cluster_selection_epsilon,
|
|
541
|
+
recursive=clustering_cfg.recursive,
|
|
542
|
+
computation_cfg=computation_cfg,
|
|
543
|
+
save_cfg=save_cfg,
|
|
544
|
+
save_labels_dir=save_labels_dir,
|
|
545
|
+
labels_fmt=labels_fmt,
|
|
546
|
+
sampling_cfg=clustering_cfg.sampling_cfg,
|
|
547
|
+
)
|
|
548
|
+
|
|
549
|
+
def _cluster(
|
|
550
|
+
self,
|
|
551
|
+
features: SimpleMatrixFeatures,
|
|
552
|
+
sorting: DARTsortSorting,
|
|
553
|
+
recording: BaseRecording,
|
|
554
|
+
motion_est=None,
|
|
555
|
+
) -> np.ndarray:
|
|
556
|
+
return cluster_util.recursive_hdbscan_clustering(
|
|
557
|
+
features.features,
|
|
558
|
+
min_cluster_size=self.min_cluster_size,
|
|
559
|
+
min_samples=self.min_samples,
|
|
560
|
+
cluster_selection_epsilon=self.cluster_selection_epsilon,
|
|
561
|
+
recursive=self.recursive,
|
|
562
|
+
)
|
|
563
|
+
|
|
564
|
+
|
|
565
|
+
class ScikitLearnClusterer(Clusterer):
|
|
566
|
+
def __init__(self, sklearn_class_name="DBSCAN", sklearn_kwargs=None, **kwargs):
|
|
567
|
+
super().__init__(**kwargs)
|
|
568
|
+
self.sklearn_class_name = sklearn_class_name
|
|
569
|
+
self.sklearn_kwargs = sklearn_kwargs or {}
|
|
570
|
+
|
|
571
|
+
@classmethod
|
|
572
|
+
def from_config(
|
|
573
|
+
cls,
|
|
574
|
+
clustering_cfg: ClusteringConfig | None,
|
|
575
|
+
computation_cfg: ComputationConfig | None = None,
|
|
576
|
+
save_cfg=None,
|
|
577
|
+
save_labels_dir=None,
|
|
578
|
+
labels_fmt=None,
|
|
579
|
+
) -> Self:
|
|
580
|
+
assert clustering_cfg is not None
|
|
581
|
+
return cls(
|
|
582
|
+
sklearn_class_name=clustering_cfg.sklearn_class_name,
|
|
583
|
+
sklearn_kwargs=clustering_cfg.sklearn_kwargs,
|
|
584
|
+
computation_cfg=computation_cfg,
|
|
585
|
+
save_cfg=save_cfg,
|
|
586
|
+
save_labels_dir=save_labels_dir,
|
|
587
|
+
labels_fmt=labels_fmt,
|
|
588
|
+
sampling_cfg=clustering_cfg.sampling_cfg,
|
|
589
|
+
)
|
|
590
|
+
|
|
591
|
+
def _cluster(
|
|
592
|
+
self,
|
|
593
|
+
features: SimpleMatrixFeatures,
|
|
594
|
+
sorting: DARTsortSorting,
|
|
595
|
+
recording: BaseRecording,
|
|
596
|
+
motion_est=None,
|
|
597
|
+
) -> np.ndarray:
|
|
598
|
+
skcls = getattr(sklearn.cluster, self.sklearn_class_name)
|
|
599
|
+
clus = skcls(**self.sklearn_kwargs)
|
|
600
|
+
return clus.fit_predict(features.features)
|
|
601
|
+
|
|
602
|
+
|
|
603
|
+
clustering_strategies["sklearn"] = ScikitLearnClusterer
|
|
604
|
+
|
|
605
|
+
|
|
606
|
+
class Refinement(Clusterer):
|
|
607
|
+
def __init__(
|
|
608
|
+
self, clusterer: Clusterer, refinement_cfg: RefinementConfig, **kwargs
|
|
609
|
+
):
|
|
610
|
+
super().__init__(**kwargs)
|
|
611
|
+
self.clusterer = clusterer
|
|
612
|
+
self.refinement_cfg = refinement_cfg
|
|
613
|
+
self.sampling_cfg = refinement_cfg.sampling_cfg
|
|
614
|
+
|
|
615
|
+
def cluster(
|
|
616
|
+
self,
|
|
617
|
+
features: SimpleMatrixFeatures,
|
|
618
|
+
sorting: DARTsortSorting,
|
|
619
|
+
recording: BaseRecording,
|
|
620
|
+
motion_est=None,
|
|
621
|
+
):
|
|
622
|
+
sorting = self.clusterer.cluster(features, sorting, recording, motion_est)
|
|
623
|
+
sorting = self.refine(features, sorting, recording, motion_est)
|
|
624
|
+
return sorting
|
|
625
|
+
|
|
626
|
+
def _refine(
|
|
627
|
+
self,
|
|
628
|
+
features: SimpleMatrixFeatures,
|
|
629
|
+
sorting: DARTsortSorting,
|
|
630
|
+
recording: BaseRecording,
|
|
631
|
+
motion_est=None,
|
|
632
|
+
):
|
|
633
|
+
del features, recording, motion_est
|
|
634
|
+
return sorting
|
|
635
|
+
|
|
636
|
+
def refine(
|
|
637
|
+
self,
|
|
638
|
+
features: SimpleMatrixFeatures,
|
|
639
|
+
sorting: DARTsortSorting,
|
|
640
|
+
recording: BaseRecording,
|
|
641
|
+
motion_est=None,
|
|
642
|
+
):
|
|
643
|
+
sorting = self._refine(features, sorting, recording, motion_est)
|
|
644
|
+
if self.labels_fmt and self.save_labels_dir is not None:
|
|
645
|
+
labels_fmt = self.labels_fmt.format(stepname="")
|
|
646
|
+
ds_save_intermediate_labels(
|
|
647
|
+
labels_fmt, sorting, self.save_labels_dir, self.save_cfg
|
|
648
|
+
)
|
|
649
|
+
return sorting
|
|
650
|
+
|
|
651
|
+
|
|
652
|
+
refinement_strategies["none"] = Refinement
|
|
653
|
+
|
|
654
|
+
|
|
655
|
+
class TMMRefinement(Refinement):
|
|
656
|
+
def _refine(
|
|
657
|
+
self,
|
|
658
|
+
features: SimpleMatrixFeatures,
|
|
659
|
+
sorting: DARTsortSorting,
|
|
660
|
+
recording: BaseRecording,
|
|
661
|
+
motion_est=None,
|
|
662
|
+
):
|
|
663
|
+
assert features is not None
|
|
664
|
+
subsampling, ixs = self.handle_sampling(features)
|
|
665
|
+
ixs = cast(np.ndarray, ixs) if subsampling else None
|
|
666
|
+
sorting = mixture.tmm_demix(
|
|
667
|
+
sorting=sorting,
|
|
668
|
+
motion_est=motion_est,
|
|
669
|
+
refinement_cfg=self.refinement_cfg,
|
|
670
|
+
computation_cfg=self.computation_cfg,
|
|
671
|
+
fit_indices=ixs,
|
|
672
|
+
save_step_labels_format=self.labels_fmt,
|
|
673
|
+
save_step_labels_dir=self.save_labels_dir,
|
|
674
|
+
save_cfg=self.save_cfg,
|
|
675
|
+
)
|
|
676
|
+
gc.collect()
|
|
677
|
+
torch.cuda.empty_cache()
|
|
678
|
+
return sorting
|
|
679
|
+
|
|
680
|
+
|
|
681
|
+
refinement_strategies["tmm"] = TMMRefinement
|
|
682
|
+
|
|
683
|
+
|
|
684
|
+
class SplitMergeRefinement(Refinement):
|
|
685
|
+
def _refine(
|
|
686
|
+
self,
|
|
687
|
+
features: SimpleMatrixFeatures,
|
|
688
|
+
sorting: DARTsortSorting,
|
|
689
|
+
recording: BaseRecording,
|
|
690
|
+
motion_est=None,
|
|
691
|
+
):
|
|
692
|
+
return refine_util.split_merge(
|
|
693
|
+
recording=recording,
|
|
694
|
+
sorting=sorting,
|
|
695
|
+
motion_est=motion_est,
|
|
696
|
+
split_cfg=self.refinement_cfg.split_cfg,
|
|
697
|
+
merge_cfg=self.refinement_cfg.merge_cfg,
|
|
698
|
+
merge_template_cfg=self.refinement_cfg.merge_template_cfg,
|
|
699
|
+
computation_cfg=self.computation_cfg,
|
|
700
|
+
)
|
|
701
|
+
|
|
702
|
+
|
|
703
|
+
# code too old. todo.
|
|
704
|
+
# refinement_strategies["splitmerge"] = SplitMergeRefinement
|
|
705
|
+
|
|
706
|
+
|
|
707
|
+
class PCMergeRefinement(Refinement):
|
|
708
|
+
def _refine(
|
|
709
|
+
self,
|
|
710
|
+
features: SimpleMatrixFeatures,
|
|
711
|
+
sorting: DARTsortSorting,
|
|
712
|
+
recording: BaseRecording,
|
|
713
|
+
motion_est=None,
|
|
714
|
+
):
|
|
715
|
+
return refine_util.pc_merge(
|
|
716
|
+
sorting,
|
|
717
|
+
refinement_cfg=self.refinement_cfg,
|
|
718
|
+
motion_est=motion_est,
|
|
719
|
+
computation_cfg=self.computation_cfg,
|
|
720
|
+
)
|
|
721
|
+
|
|
722
|
+
|
|
723
|
+
refinement_strategies["pcmerge"] = PCMergeRefinement
|
|
724
|
+
|
|
725
|
+
|
|
726
|
+
class ForwardBackwardEnsembler(Refinement):
|
|
727
|
+
"""If there are more time chunk ones, make a new ABC with this logic."""
|
|
728
|
+
|
|
729
|
+
def cluster(
|
|
730
|
+
self,
|
|
731
|
+
features: SimpleMatrixFeatures,
|
|
732
|
+
sorting: DARTsortSorting,
|
|
733
|
+
recording: BaseRecording,
|
|
734
|
+
motion_est=None,
|
|
735
|
+
):
|
|
736
|
+
chunk_length_samples = (
|
|
737
|
+
recording.sampling_frequency * self.refinement_cfg.chunk_size_s
|
|
738
|
+
)
|
|
739
|
+
chunk_time_ranges_s = chunk_time_ranges(recording, chunk_length_samples)
|
|
740
|
+
times_seconds = sorting.times_seconds # type: ignore
|
|
741
|
+
assert features is not None
|
|
742
|
+
|
|
743
|
+
chunk_sortings = []
|
|
744
|
+
for lo, hi in chunk_time_ranges_s:
|
|
745
|
+
mask = np.flatnonzero(times_seconds == times_seconds.clip(lo, hi))
|
|
746
|
+
s = sorting.mask(mask)
|
|
747
|
+
f = features.mask(mask)
|
|
748
|
+
l = self.clusterer._cluster(f, s, recording, motion_est)
|
|
749
|
+
labels = np.full_like(sorting.labels, -1)
|
|
750
|
+
labels[mask] = l
|
|
751
|
+
chunk_sortings.append(sorting.ephemeral_replace(labels=labels))
|
|
752
|
+
|
|
753
|
+
labels = forward_backward.forward_backward(
|
|
754
|
+
chunk_time_ranges_s,
|
|
755
|
+
chunk_sortings,
|
|
756
|
+
log_c=self.refinement_cfg.log_c,
|
|
757
|
+
feature_scales=self.refinement_cfg.feature_scales,
|
|
758
|
+
adaptive_feature_scales=self.refinement_cfg.adaptive_feature_scales,
|
|
759
|
+
motion_est=motion_est,
|
|
760
|
+
)
|
|
761
|
+
sorting = sorting.ephemeral_replace(labels=labels)
|
|
762
|
+
return sorting
|
|
763
|
+
|
|
764
|
+
|
|
765
|
+
refinement_strategies["forwardbackward"] = ForwardBackwardEnsembler
|