Scott Linderman
Scott Linderman
This is a rather vague issue... I'm working on a few projects that have very large and very sparse observations. That motivated the development of the `ZeroInflatedCountModel`s, which takes in...
most of the plumbing is there, but we need to update `_set_expected_statistics` to provide input statistics. we also need tests for these changes. see the `laplace` branch.
Upgrading code to support Python3 standards. Automatically done with 2to3. See corresponding issue: https://github.com/mattjj/svae/issues/5
I will do it.
Next time, make sure to show the classic mode-seeking vs mode-covering figure for KL(q||p) vs KL(p||q)!
I updated to Numpy 2.0 and found that the `MultivariateNormalDiag` and `MultivariateNormalFullCovariance` constructors crashed because `np.issctype` has been removed. Is Numpy 2.0 supported, or will it be soon? Here is...
Tensorflow probability broke for numpy 2.0. See https://github.com/tensorflow/probability/issues/1814. Once that issue is resolved, we can unpin the numpy
The iterated smoothers should iteratively perform a forward/backward pass, using linearizations around the previous iteration's smoothed trajectory. Sarkka (2023), Ch 13 describes the algorithm well. Our implementations are only iterating...
If we drop the Kilosort lab, we should add the conv and cross correlation sections back to another lab. Otherwise the students have to figure out how to use convs...
The 2025 version of the course has M/W lectures, but that hasn't allowed enough time to get into the mathematical derivations of the algorithms. I had to rush the coordinate...