fix(dyn/synapses): DualExp equal-tau + STP per-neuron reset

[chaoming0625]

Fresh review of brainpy/dyn/{synapses,projections}. Three High fixes:

• DualExpon raised ZeroDivisionError/NaN when tau_rise == tau_decay; now uses the e/tau limit element-wise (matches Alpha).
• DualExponV2 silently produced zeros/NaN for equal taus; now raises a clear ValueError pointing to bp.dyn.Alpha.
• STP.reset_state crashed for a per-neuron array U (Variable.fill_ needs a scalar); now broadcasts.

Prior-audit synapse/projection bugs (STP u growth, double-comm, STDP, PoissonInput std) verified already-fixed. In-scope: 181 passed. Findings: docs/issues-found-20260619-dyn-synapses.md.

Summary by Sourcery

Fix edge-case behaviour of dual-exponential synapses and STP reset, and record audit findings for dyn/synapses and projections.

Bug Fixes:

• Handle DualExpon synapses with equal rise and decay time constants without crashes or NaNs by using a well-defined alpha-function limit.
• Make DualExponV2 explicitly reject equal-tau configurations with a clear error message instead of producing invalid or silent-zero outputs.
• Allow STP synapses to reset correctly when the release probability U is provided as a per-neuron array, including in batched modes.

Documentation:

• Add an audit report documenting identified and previously fixed issues in dyn/synapses and dyn/projections, including rationale for recorded-but-unfixed items.

Tests:

• Add regression tests covering DualExpon equal-tau behaviour, DualExponV2 equal-tau error handling, and STP reset and dynamics with scalar and per-neuron U configurations.

[sourcery-ai]

Reviewer's Guide

Fixes equal time-constant handling for dual-exponential synapses and makes STP reset support per-neuron U, plus adds regression tests and an audit findings doc for dyn/synapses and projections.

File-Level Changes

Change	Details	Files
Handle DualExpon equal tau_rise == tau_decay robustly and clarify DualExponV2 behaviour.	Refactored computation of the DualExpon input-scaling coefficient a into a dedicated helper that uses an algebraically stable closed form and an e/tau L’Hôpital limit for equal taus. Updated DualExpon.init to use the new helper instead of the previous A-based expression that suffered from division-by-zero and NaN when tau_rise == tau_decay. Extended the dual-exponential peak-normalizer helper to detect tau_rise == tau_decay when A is auto-computed and raise a clear ValueError explaining that DualExponV2 is singular in that regime and pointing users to Alpha or explicit A.	brainpy/dyn/synapses/abstract_models.py
Fix STP.reset_state to support per-neuron array U while preserving batched behaviour.	Changed STP.reset_state to initialise u and then broadcast-multiply by U instead of using Variable.fill_, which only accepts scalar fills. Ensured the new initialisation works both for scalar U and heterogeneous per-neuron U in batched and non-batched modes.	brainpy/dyn/synapses/abstract_models.py
Add regression tests for DualExpon equal-tau behaviour and STP per-neuron U, and document the broader synapse/projection audit.	Added unit tests that cover DualExpon equal-tau construction, numerical behaviour matching a normalized alpha function, and DualExponV2 raising a clear ValueError for equal taus. Added unit tests verifying that STP supports array-valued U at reset, integrates without NaNs for heterogeneous U, and preserves existing scalar-U batched behaviour. Added a markdown document capturing the 2026-06-19 dyn/synapses and dyn/projections audit results, including both fixed and recorded-only issues.	brainpy/dyn/synapses/abstract_models_test.py docs/issues-found-20260619-dyn-synapses.md

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Hey - I've left some high level feedback:

• The exact equality check self.tau_rise == self.tau_decay in _dual_exp_a (and _format_dual_exp_A) may behave unintuitively for non-integer or JIT-transformed parameters; consider using a small tolerance (or documenting that only exactly equal taus trigger the alpha-limit behavior/error) to make the edge case more predictable.
• In STP.reset_state, the broadcasting via self.u.value = self.u.value * self.U works but is a bit opaque; using an explicit broadcast (e.g., bm.broadcast_to with the target shape or a helper that mirrors init_variable) would make the intended semantics with batched modes and heterogeneous U clearer and less error-prone if shapes change later.
• The new _dual_exp_a helper partially duplicates the logic in _format_dual_exp_A and relies on the caller to know when each path is valid; consider centralizing the dual-exp normalization (including the equal-tau handling and L’Hôpital limit) into a single, well-documented function to avoid future divergence between DualExpon and DualExponV2.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- The exact equality check `self.tau_rise == self.tau_decay` in `_dual_exp_a` (and `_format_dual_exp_A`) may behave unintuitively for non-integer or JIT-transformed parameters; consider using a small tolerance (or documenting that only *exactly* equal taus trigger the alpha-limit behavior/error) to make the edge case more predictable.
- In `STP.reset_state`, the broadcasting via `self.u.value = self.u.value * self.U` works but is a bit opaque; using an explicit broadcast (e.g., `bm.broadcast_to` with the target shape or a helper that mirrors `init_variable`) would make the intended semantics with batched modes and heterogeneous `U` clearer and less error-prone if shapes change later.
- The new `_dual_exp_a` helper partially duplicates the logic in `_format_dual_exp_A` and relies on the caller to know when each path is valid; consider centralizing the dual-exp normalization (including the equal-tau handling and L’Hôpital limit) into a single, well-documented function to avoid future divergence between `DualExpon` and `DualExponV2`.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}