paulfharrison.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

Comparison of optimization and sampling from a distribution defined by an energy function. I use a continuous version of the Ising model spin lattice energy. First, optimization from a random initial state using gradient descent with momentum, using the SGD optimizer in PyTorch.

submitted 5 days ago • 1 comment

I think I'm a little bit in love with the withr package in R. Similar to "with" in Python, you can guarantee to properly clean up when using a resource such as a connection to a file, a temporary directory, or a temporary global setting change. withr.r-lib.org

submitted 15 days ago • 1 comment

Bash Strict Mode: Stop Silent Failures in Your Scripts 🧵 1/ Ever had a Bash script fail silently, leading to hours of debugging? Bash doesn't fail by default—it keeps running even after errors. Here's how to fix it with strict mode. 👇

submitted 17 days ago • 1 comment

I've been playing with a Langevin Dynamics sampler as an optimizer in PyTorch -- SGD with momentum and noise. It's a remarkably simple algorithm for sampling from a distribution (approximately). Should eliminate over-fitting, and can also be used for generative tasks.

submitted 17 days ago • 1 comment

Trying to bring some old family photos back to life with g.i.m.p and its truly amazing resynthesizer add-on created by @paulfharrison.bsky.social getting some pretty good results. Takes me a few hours per photo, but is super satisfying…

submitted 38 days ago • 0 comments

This is long overdue, but over the winter break we were finally able to write up our sgdGMF paper: arxiv.org/abs/2412.20509 We present a stochastic gradient descent method that allows to efficiently and very quickly estimate latent factors for, e.g., dimensionality reduction of single-cell data

submitted 37 days ago • 3 comments

If I were to do some virtual half-day workshops this year, what topics and depths might be useful?

submitted 52 days ago • 3 comments

I've been watching the 2023 Statistical Rethinking lecture series by Richard McElreath. These cover a complete approach to statistics based on causal reasoning and Bayesian analysis. They are excellent, highly recommend. www.youtube.com/playlist?lis...

submitted 64 days ago • 2 comments

☎️ calling #SingleCell researchers in 🇦🇺! Join our last community meeting for the year Hear from @lgmartelotto.bsky.social and @swbioinf.bsky.social and chat all things single cell or #SpatialOmics! www.biocommons.org.au/events/sc-sp... #bioinformatics #lifescience

submitted 84 days ago • 0 comments

Some years ago I wrote some code that produced a bad plot. I then had to endure multiple people having to explain my bad plot in seminars.

submitted 84 days ago • 2 comments

Starting today, The Carpentries will be active on BlueSky! 🥳 We'll continue to maintain our Mastodon (hachyderm.io/@thecarpentr...) and our LinkedIn (www.linkedin.com/company/the-...) accounts, and the content we post on here will be the same as the content we post on Mastodon. Excited to be here!

submitted 87 days ago • 2 comments

I just ran my 1.5 hour HTML+SVG+JavaScript+D3 workshop at ResBaz Victoria 2024. Touched on hopefully most of the important ideas needed for an interactive web-based data visualization. #resbaz #resbazvic #resbazvic2024 pfh.github.io/jsfoot/

submitted 93 days ago • 1 comment

Our paper "A Plot is Worth a Thousand Tests: Assessing Residual Diagnostics with the Lineup Protocol" has appeared in issue 3 of JCGS, www.tandfonline.com/doi/full/10.... . It is open access for now, but if you ever have problems accessing you'll find the preprint at doi.org/10.48550/arX...

submitted 98 days ago • 0 comments

I've begun a statistical graphics starter pack at go.bsky.app/EMb8xWf . Please help me find more people or feeds to add to this pack. #rstats #visualisation #statistics 📈

submitted 100 days ago • 26 comments

Pondering FDR control. Procedures like BH represent a certain policy, which can be confirmed to work by simulation for any given number of true discoveries. The actually achieved FDR guarantee is the FDR obtained in the worst case.

submitted 101 days ago • 1 comment

Our collaborative work with Narry Kim on the kinetics of mRNA poly(A) tail shortening. Led by Young-suk Lee, who established his own lab while the work was in progress, and Yevgen Levdansky in my lab. #RNASky #RNABiology www.nature.com/articles/s41...

submitted 374 days ago • 2 comments

Something that should exist: geom_principal_curve() If x and y have a symmetric relationship, eg both x and y are noisy measurements of an underlying hidden variable, geom_smooth will under-estimate the slope, just like linear regression.

submitted 104 days ago • 1 comment

sgdGMF is a general purpose matrix factorization package. Like PCA, but better, e.g. works well with count data. Should be fast, e.g. suitable for scRNA-Seq. Presented by @davi1893.bsky.social at #abacbs2024. github.com/CristianCast...

submitted 112 days ago • 0 comments

The Knockoff Framework looks very interesting as a broadly applicable method of maintaining an FDR. (Guannan Yang talked about this at #abacbs2024) web.stanford.edu/group/candes...

submitted 113 days ago • 0 comments

Jingyi Jessica Li demonstrates how to use permutations to test the accuracy of tSNE and UMAP #ABACBS2024

submitted 115 days ago • 0 comments

A nice video on power calculation by simulation. www.youtube.com/watch?v=vE8b...

submitted 124 days ago • 0 comments

Registrations for ResBas Victoria 2024 at Monash University are now open. The program is looking quite varied, it's going to be an adventure! #resbaz #resbazvic2024 resbaz.github.io/resbazvic202...

submitted 126 days ago • 0 comments

New CRAN Task View: Dynamic Visualizations and Interactive Graphics #rstats #dataviz By Sherry Zhang, Dianne Cook @visnut.bsky.social URL: cran.r-project.org/view=Dynamic...

submitted 155 days ago • 1 comment

High FDR rates from edgeR and DESeq2 for large datasets, seems a little concerning. I have noticed edgeR has a tendency to call genes differentially expressed on the basis of a single sample with high expression (have not tried with latest version). genomebiology.biomedcentral.com/articles/10....

submitted 148 days ago • 1 comment

This week the Monash Genomics and Bioinformatics Platform did a bulk RNA-Seq workshop, covering end-to-end from experimental design, through library preparations, running an analysis pipeline, and digging into differential expression.

submitted 148 days ago • 1 comment