Michael Shvartsman

Make lookahead acquisition compatible with posterior transforms

4

Summary: More BoTorch API compatibility fixes, this time for lookahead LSE acquisition. Differential Revision: D36333559

Contrast sensitivity example

1

Summary: This is a full example experiment with psychopy, measuring the CSF in up to 8d. Differential Revision: D37896950

Support ragged no-op configs for benchmark

2

Summary: This is useful when some benchmark configurations are invalid Differential Revision: D36214512

[matnormal] Matrix-normal (dual probabilistic) SRM

4

This is a TF2 port of the matrix-normal SRM code we used in https://arxiv.org/abs/1711.03058. It's not quite ready for merge but I'm putting up the PR to start discussion and...

hyperparamopt: Standardize hyperparameter optimization

2

Most of the methods in brainiak have hyperparameters that need to be tuned. It would be useful to make hyperparameter optimization easy for practitioners to apply. There have been a...

A lot of the structure-exploiting linear algebra in `brainiak.matnormal` is basically the same as in `tf.linalg.LinearOperator`. I'm not sure if I didn't know about it or it didn't exist in...

Looking at Anderson et al. 2016 (https://arxiv.org/pdf/1608.04647.pdf) expression 13 the 2nd term $\Sigma_{\mu}$ is not indexed by $k$. On the other hand, in the code here https://github.com/brainiak/brainiak/blob/master/brainiak/factoranalysis/tfa.py#L725 `template_centers_mean_cov` (which I...

In BRSA/GBRSA the `score` and `transform` functions use assumptions on X and X0 that do not exist in the fit method (namely that both X and X0 are AR1). Additionally,...

Expose logp for Bayesian models in brainiak

4

SRM and (H)TFA, and Bayesian RSA are all Bayesian models. It would be great if they exposed a method that produces a log-likelihood given some data and parameters. This would...

[Bug] Possible memory leak in `botorch.optim.optimize_acqf`

13

# 🐛 Bug As far as I can tell `botorch.optim.optimize_acqf` leaves a tiny bit of memory behind somewhere. It seems worse for q-batched acquisition functions (at least, for qUCB and...