Skip to content

Multicompartment Connection

Latest
Compare
Choose a tag to compare
@Hananel-Hazan Hananel-Hazan released this 18 Oct 17:36
· 16 commits to master since this release
1b97861

Here's a clearer and more friendly version of the text:


What's New

In this version, along with our usual performance improvements (thanks to @C-Earl and @SimonInParis!), we've introduced the MulticompartmentConnection class for more complex connections. This allows spikes to pass through different features like weights, biases, and boolean masks in a specified order. You can create custom pipelines with these features, and they’ll execute in the order you define. For example, you can use a pipeline that first multiplies spikes by weights, then adds a bias. You can even add more features before, after, or between these steps!

For more details, check out our documentation.

General Updates:

  • Added MulticompartmentConnection #695 by @C-Earl
  • Various package updates and reformatting #542
  • Python updated to 3.11 #550
  • PyTorch updated to 2.0 #630
  • Maintenance updates, dependency bumps, and more! #593, #602

Bug Fixes:

  • Fixed an issue with net.run for unconnected layers #547
  • Corrected plotting aspect ratios #651
  • Fixed small bugs in environment_pipeline #579
  • Addressed typo errors #661

Performance Enhancements:

  • Added vector learning rates for all UpdateRules #552
  • CUDA updates and performance fixes #655

Code Refactoring:

  • Made the code more "Pythonic" #570
  • Removed debug print statements #566

Examples & Documentation:

  • Updated mnist examples #557
  • Addressed issues with Read the Docs #642

Dependency Updates:

  • Bumped various dependencies, including jupyter-server, pillow, requests, and more.

New Contributors

A warm welcome to our new contributors:

Full Changelog: Compare changes from 0.3.1 to 0.3.3


This version organizes the updates to help you quickly find what’s new and improved. Enjoy the new features, and as always, thanks for your contributions!