Jens Egholm Pedersen

That's a great fix @realPtolemy ! Thank you for presenting it. Would you be up for writing that in a PR? I think that would help a lot of other...

Good questions. The neuron size of 1 is definitely not by choice. I think the issue comes from the fact that time constants are typically given as a single number....

Can you share details for reproduction?

Here's a test that currently fails. Saving works, but loading fails with an `EOFError` ```python def test_serialize(): import tempfile s = MockState(torch.rand(10)) tf = tempfile.NamedTemporaryFile() torch.save(s, tf) new_s = torch.load(tf,...

You bring an excellent point. I think it's fair to say that we haven't done a good job of linking the continuous-time formulation (such as $f \star g$ for the...

> > The output is indeed the result of the operation and the formalization closely follows the parameters for the [PyTorch convolutions](https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html). > > Ah OK, good to know about...

> From what I have seen so far, most components need to be able to accept both spike and matrix inputs depending on the output of the component before them,...

In principle this looks great. You're adding an attribute that's definitely reasonable for many of the models. Simultaneously, this deviates from the neuron models, which means that all the platforms...

I think @stevenabreu7 brings up an excellent point which I totally overlooked. NIR is inherently continuous, and in a continuous setting the reset is instantaneous. From that perspective it might...

Fantastic discussion here. Thanks everyone for mulling this over so carefully. And thank you for adding backwards compatible code @benkroehs! I like the separation of concerns you made, @bvogginger. Let...