Release 0.3.5

.github/PULL_REQUEST_TEMPLATE.md

@@ -15,8 +15,8 @@
 - [ ] Does this change simulated results (spikes, voltages, traces, weights)?
       - [ ] No.
       - [ ] Yes. I regenerated the golden files in `test/regression/` with
-            `--reason`, and explain below why the results changed. (Results-changing
-            fixes are released in 0.4.0, not 0.3.x; see `CONTRIBUTING.md`.)
+            `--reason`, and explain below why the results changed. Add a
+            "Results changes" entry to `CHANGELOG.md`; see `CONTRIBUTING.md`.)
 - [ ] Did any of these files change? If yes, say which and why:
       - golden files under `test/regression/`
       - operation-count file under `test/regression/` (if present)

CHANGELOG.md

@@ -6,6 +6,8 @@ see the [GitHub releases / tags](https://github.com/BindsNET/bindsnet/releases).
 
 ## [Unreleased]
 
+## [0.3.5] - 2026-10-29
+
 ### Results changes (read this before upgrading)
 These fixes change numbers that code produced before. Everything else in this
 release either was a crash before or gives bit-identical results (checked by the
@@ -147,6 +149,9 @@ golden-output tests in `test/regression/`).
 - `test/environment/test_gym_environment.py`.
 
 ### Changed
+- Release rule (`CONTRIBUTING.md`): fixes that change results may ship in any
+  release and are always listed first, under "Results changes". Before, they waited
+  for 0.4.0. Changes that break user code still wait for the next minor release.
 - Convolutional and locally connected connections normalize all kernels in one
   operation instead of a Python loop (6 to 45 times faster on CPU). Results are
   identical on CPU; on CUDA, float32 kernels of 81 or more weights can differ by

CITATION.cff

@@ -2,7 +2,7 @@ cff-version: 1.2.0
 message: "If you use BindsNET, please cite the article below."
 title: "BindsNET: A Machine Learning-Oriented Spiking Neural Networks Library in Python"
 type: software
-version: 0.3.4
+version: 0.3.5
 license: AGPL-3.0-only
 doi: 10.5281/zenodo.20695115
 repository-code: "https://github.com/BindsNET/bindsnet"

CONTRIBUTING.md

@@ -110,10 +110,12 @@ harmless. Please do not edit them by hand to make them pass.
   head takes more than 1.25 times as long (median over the rounds) and the lower end of
   the 95% bootstrap interval of that ratio is above 1. It is not a required check; shared CI machines are noisy, and a maintainer decides.
 
-**Fixes that change results are released in 0.4.0, not in 0.3.x.** A correct fix that
-changes what a simulation produces still changes published numbers for people using
-0.3.x, so it waits for the next minor release. State in the pull request whether your
-change alters results, and add a `CHANGELOG.md` entry that says so.
+**Fixes that change results can ship in any release, and are always announced.** A
+correct fix that changes what a simulation produces still changes numbers people have
+published, so `CHANGELOG.md` lists it first, under "Results changes", naming the affected
+classes or models. State in the pull request whether your change alters results, and add
+that entry. Changes that break user code (a removed or renamed argument, a different
+return type) wait for the next minor release (0.4.0).
 
 ## Releasing to PyPI
 

docs/source/models_spec.rst

@@ -6,7 +6,7 @@ models BindsNET implements: the difference equations actually solved each timest
 the default parameters with units. Equations and defaults below were transcribed from
 the source (``bindsnet/network/nodes.py`` and ``bindsnet/learning/learning.py``); when
 in doubt, the source is authoritative. Parameter defaults are stated as of package
-version 0.3.4.
+version 0.3.5.
 
 Discretization
 --------------

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "bindsnet"
-version = "0.3.4"
+version = "0.3.5"
 description = "Spiking neural networks for ML in Python"
 authors = [
     { name = "Hananel Hazan", email = "hananel@hazan.org.il" },

test/regression/README.md

@@ -53,9 +53,9 @@ Your change altered a result. Either that is the point of the change, or it is a
 python test/regression/regen.py --reason "why the results changed"
 ```
 
-The script refuses to run without a reason and stores it in every file. Results-changing
-fixes are released in version 0.4.0, not in a 0.3.x patch release, so the pull request
-will wait for that release. See `CONTRIBUTING.md`, "Tests that guard results".
+The script refuses to run without a reason and stores it in every file. A results-changing
+fix also needs an entry under "Results changes" in `CHANGELOG.md` that names the affected
+classes or models. See `CONTRIBUTING.md`, "Tests that guard results".
 
 ## When an operation-count test fails