Nengo 2.1 released

The Nengo team is happy to announce the release of Nengo 2.1.

What is Nengo?

Nengo is a Python library for building and simulating large-scale neural models for AI and robotics. It can be thought of as a neural compiler, transforming a functional description of a neural model to a network of spiking or non-spiking neurons that can run on multiple backends including GPUs and neuromorphic hardware.

What’s new?

Since Nengo 2.0, the ecosystem has grown significantly. We have recently released a graphical interface called Nengo GUI to look at different parts of your model as it simulates, and Nengo OCL to speed up large simulations by taking advantage of your GPU.

Nengo itself has learned several new features. Some highlights:

  • The output of a node can now be properly reset by using a Process. Processes can also be used to inject noise into an ensemble.
  • We added a new learning rule called Voja that updates an ensemble’s encoders to fire selectively to its inputs.

We also changed how learning rules get error information. Instead of the Connection.modulatory attribute, you provide error information by connecting to the learning rule. Additionally, if you were probing a connections’ 'decoders' or 'transform' to see how weights were changing during learning, you should now probe the more general 'weights'.

To see the full list of changes in Nengo 2.1, head to the Github release page.

How do I get it?

To get the new version of Nengo, use pip.

pip install --upgrade nengo

Where can I learn more?