Add SMC-PHD components #798

sglvladi · 2023-05-02T14:51:59Z

This PR adds the following components:

SMCPHDPredictor: Performs the SMC-PHD prediction step
SMCPHDUpdater: Performs the SMC-PHD update step
SimpleHypothesiser: A configurable hypothesiser that generates a sequence of SingleHypothesis objects (used to generate the hypotheses required by the SMCPHDUpdater)

A running example can be found here

codecov · 2024-02-15T16:23:37Z

Codecov Report

Attention: Patch coverage is 94.69027% with 6 lines in your changes are missing coverage. Please review.

Project coverage is 93.42%. Comparing base (cd30860) to head (27ecd5b).
Report is 30 commits behind head on main.

Files	Patch %	Lines
stonesoup/updater/particle.py	87.50%	2 Missing and 3 partials ⚠️
stonesoup/hypothesiser/simple.py	97.05%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #798      +/-   ##
==========================================
+ Coverage   93.40%   93.42%   +0.01%     
==========================================
  Files         201      202       +1     
  Lines       12700    12813     +113     
  Branches     2608     2621      +13     
==========================================
+ Hits        11863    11970     +107     
- Misses        593      595       +2     
- Partials      244      248       +4

Flag	Coverage Δ
integration	`66.15% <36.28%> (-0.26%)`	⬇️
unittests	`89.04% <94.69%> (+0.05%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

…cleFilter and tidy up tests

jswright-dstl

Looks good. Some spelling mistakes and spelling consistency changes.

jswright-dstl · 2024-02-19T11:41:59Z

stonesoup/updater/particle.py

+    .. note::
+
+        - It is assumed that the proposal distribution is the same as the dynamics
+        - Target "spawing" is not implemented


Suggested change

- Target "spawing" is not implemented

- Target "spawning" is not implemented

jswright-dstl · 2024-02-19T12:05:22Z

stonesoup/updater/particle.py

+
+        # Resample
+        if self.resampler is not None:
+            # Normalize weights


Suggested change

# Normalize weights

# Normalise weights

jswright-dstl · 2024-02-19T12:05:34Z

stonesoup/updater/particle.py

+            updated_state.log_weight = log_post_weights - log_num_targets
+            # Resample
+            updated_state = self.resampler.resample(updated_state, self.num_samples)
+            # De-normalize


Suggested change

# De-normalize

# De-normalise

sdhiscocks · 2024-02-21T15:26:37Z

stonesoup/hypothesiser/simple.py

+
+            # Compute measurement prediction
+            if self.predict_measurement:
+                measurement_prediction = self.updater.predict_measurement(


Codecov reported as not covered by test, which seems odd when test below should be setting predict_measurement to True and False.

Nice spot. I have added an extra test case to make sure we reach that part of the code.

sdhiscocks · 2024-02-21T16:15:00Z

stonesoup/predictor/particle.py

+    .. note::
+
+        - It is assumed that the proposal distribution is the same as the dynamics
+        - Target "spawing" is not implemented


Suggested change

- Target "spawing" is not implemented

- Target "spawning" is not implemented

sdhiscocks · 2024-02-23T10:26:31Z

stonesoup/predictor/particle.py

+        doc="The field name of the number of samples parameter for the birth sampler. This is "
+            "required since the number of samples required for the birth sampler may be vary"
+            "between iterations. Default is 'num_samples'")
+    birth_scheme: Literal["expansion", "mixture"] = Property(


I think this would be better as an Enum, but also support strings. Similar to plotting which use Enum but also supports int's.

Sounds sensible. My only question is then where to place/document this Enum.

In this file would be fine?

sdhiscocks · 2024-02-23T10:31:07Z

stonesoup/updater/particle.py

+        g = self._get_measurement_loglikelihoods(prediction, detections)
+
+        # Calculate w^{n,i} Eq. (20) of [2]
+        Ck = self.prob_detect.log() + g + prediction.log_weight[:, np.newaxis]


Suggested change

Ck = self.prob_detect.log() + g + prediction.log_weight[:, np.newaxis]

Ck = np.log(self.prob_detect) + g + prediction.log_weight[:, np.newaxis]

sdhiscocks · 2024-02-23T12:56:28Z

stonesoup/updater/particle.py

+    clutter_intensity: float = Property(
+        doc="Average number of clutter measurements per time step, per unit volume")
+    num_samples: int = Property(
+        default=1024,


Would it make sense to have option of None as well (possibly even default), where it is the set to number of particles on input prediction?

Sounds sensible. I have done this, but also added a note to emphasise that users should set this value when used in conjunction with a predictor that performs birth via expansion.

sdhiscocks · 2024-02-23T13:20:38Z

stonesoup/hypothesiser/simple.py

+
+        if self.check_timestamp:
+            # Check to make sure all detections are obtained from the same time
+            timestamps = set([detection.timestamp for detection in detections])


Suggested change

timestamps = set([detection.timestamp for detection in detections])

timestamps = {detection.timestamp for detection in detections}

…set construction

sglvladi · 2024-02-26T16:38:56Z

@sdhiscocks @jswright-dstl, I have applied the requested changes and improved the documentation slightly.

sglvladi requested a review from a team as a code owner May 2, 2023 14:52

sglvladi requested review from jswright-dstl and mharris-dstl and removed request for a team May 2, 2023 14:52

sglvladi force-pushed the smcphd branch from c84a78a to 7e80124 Compare May 2, 2023 15:15

nperree-dstl self-requested a review May 4, 2023 14:35

sglvladi requested review from sdhiscocks and removed request for mharris-dstl May 9, 2023 15:22

nperree-dstl removed their request for review June 23, 2023 09:01

sglvladi force-pushed the smcphd branch 2 times, most recently from 620ccc2 to c9d980c Compare October 19, 2023 10:07

sglvladi added 5 commits February 15, 2024 15:56

Add SMCPHDPredictor

4f3800a

Add SMCPHDUpdater

50e0b09

Add SimpleHypothesiser

e879550

Documentation updates

d50d76d

Refactor resampling code in SMCPHDUpdater

e0c73d4

sglvladi force-pushed the smcphd branch from c9d980c to e0c73d4 Compare February 15, 2024 15:57

Ensure constraint_func is implemented in SMCPHDUpdater

b7c3139

sglvladi added 2 commits February 16, 2024 01:17

Update the SMCPHDPredictor to use similar properties to BernouliParti…

7725508

…cleFilter and tidy up tests

Minor refactoring/tidy up of SMCPHDUpdater

f173b07

sglvladi force-pushed the smcphd branch from acad104 to f173b07 Compare February 16, 2024 01:18

jswright-dstl approved these changes Feb 19, 2024

View reviewed changes

sdhiscocks reviewed Feb 23, 2024

View reviewed changes

sglvladi added 3 commits February 23, 2024 14:23

Fix SimpleHypothesiser tests coverage and minor change in timestamps …

2f5c6ef

…set construction

Add SMCPHDBirthSchemeEnum and improve SMCPHDPredictor docs

346cf91

Fix typos and improve documentation of SMCPHDUpdater

27ecd5b

sglvladi force-pushed the smcphd branch from 7074e19 to 27ecd5b Compare February 26, 2024 16:12

spike-dstl approved these changes Feb 28, 2024

View reviewed changes

sdhiscocks approved these changes Feb 28, 2024

View reviewed changes

sdhiscocks merged commit 10e26e0 into dstl:main Feb 28, 2024
10 checks passed

sdhiscocks added the enhancement label May 28, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add SMC-PHD components #798

Add SMC-PHD components #798

sglvladi commented May 2, 2023

codecov bot commented Feb 15, 2024 •

edited

Loading

jswright-dstl left a comment

jswright-dstl Feb 19, 2024

jswright-dstl Feb 19, 2024

jswright-dstl Feb 19, 2024

sdhiscocks Feb 21, 2024

sglvladi Feb 26, 2024

sdhiscocks Feb 21, 2024

sdhiscocks Feb 23, 2024

sglvladi Feb 23, 2024 •

edited

Loading

sdhiscocks Feb 26, 2024

sdhiscocks Feb 23, 2024

sdhiscocks Feb 23, 2024

sglvladi Feb 26, 2024

sdhiscocks Feb 23, 2024

sglvladi commented Feb 26, 2024

	- Target "spawing" is not implemented
	- Target "spawning" is not implemented

	Ck = self.prob_detect.log() + g + prediction.log_weight[:, np.newaxis]
	Ck = np.log(self.prob_detect) + g + prediction.log_weight[:, np.newaxis]

	timestamps = set([detection.timestamp for detection in detections])
	timestamps = {detection.timestamp for detection in detections}

Add SMC-PHD components #798

Add SMC-PHD components #798

Conversation

sglvladi commented May 2, 2023

codecov bot commented Feb 15, 2024 • edited Loading

Codecov Report

jswright-dstl left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sglvladi Feb 23, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sglvladi commented Feb 26, 2024

codecov bot commented Feb 15, 2024 •

edited

Loading

sglvladi Feb 23, 2024 •

edited

Loading