Will Townes

As in GLM it would be nice to be able to include offset terms to adjust for different exposure time for Poisson variables. Ideally there would be a separate offset...

Simulations of super bowl pool

1

Hi Drew, I was inspired by your blog post about finding the probabilities of the different digit combinations. I wanted to figure out if there was an advantage to selecting...

Variational Gaussian process loss- possible math error

1

In the documentation for variational Gaussian process applied to minibatches (https://github.com/tensorflow/probability/blob/v0.12.1/tensorflow_probability/python/distributions/variational_gaussian_process.py#L572), the KL term is rescaled by batch_size/num_training_points_. I assume the reconstruction error term (expected log-likelihood) is not scaled and...

Does kallisto bus discard reads with low-quality barcodes?

3

I am trying to re-process the [data from Grun et al 2016](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2142263) (a CEL-Seq paper), and I am getting very low counts from kallisto and bustools, which are inconsistent with...

Disk-based data

4

It would be awesome to be able to run sctransform on data that are too big to fit in memory (eg, HDF5 files accessed through something like [DelayedArray](https://bioconductor.org/packages/release/bioc/html/HDF5Array.html)). Is this...

Return factors and loadings as matrices instead of data frames

2

As pointed out by https://github.com/willtownes/scrna2019/issues/5 glmpca is returning factors as a data frame instead of a matrix (as the documentation claims), which causes downstream problems converting SingleCellExperiment objects to Seurat....

Reduce glmpca_family memory consumption

1

The glmpca_family object consumes a lot of memory because it preserves the environment where `infograd` and `dev_func` functions are defined. Refactor this code so these functions depend only on their...

improve negative binomial overdispersion estimation

3

The [glmGamPoi package](https://github.com/const-ae/glmGamPoi) has some cool strategies for estimating the negative binomial overdispersion parameter(s). Try to use those ideas to improve accuracy, numerical stability, and speed of nb estimation in...

As a user, I can request that glmpca not try to learn the negative binomial dispersion parameter during optimization by setting a flag in the control parameters. This issue is...

Suggested by Aaron Lun in kstreet13/scry#7 > There's probably some low-hanging fruit with BiocParallelization and DelayedArraying, especially if you can make use of the parallelized crossprod provided by DelayedArray (via...