Brant Peterson

some damn piece of research code that isn't done failing yet - @fdmts

[Brant Peterson @fixedbydrift]

@obogobo actually just give me back giant scrolling ansi art connect screens and I'm IN

[Brant Peterson @fixedbydrift]

...or like web0.0.1a? Mastodon is bbs+fidomail change my mind

[JΔcksøn W @obogobo]

Web 2.0 is dead long live Web 1.0

[Brant Peterson @fixedbydrift]

Teamwork makes the dream work :)
Very cool result -- congratulations @nlbedford @EvolEmpiricist and the whole crew!

[Brant Peterson @fixedbydrift]

@cripple_vs_stem Ovation https://t.co/fP6b3XnhKz pipettes saved my wrists (and probably my PhD) ~15 years ago and I still remember how starkly different not-pain felt after years with R-Gs

👏 design for humans 👏

[Brant Peterson @fixedbydrift]

Congrats @ekdelaney making the world a little more color-optimal one (bag of) neural networks at a time!

There's a "stoked" joke in here somewhere...

[Kevin K. Yang 楊凱筌 @KevinKaichuang]

Comparisons surrogate functions for Bayesian optimization of protein function (with wet-lab!). Ensembles of NNs outperform GPs.

@gruver_nate @samscub @polkirichenko @m_finzi @philmaffettone @vivek_myers @ekdelaney @peytongreenside @andrewgwils

https://t.co/Im5j3RDedy

[Brant Peterson @fixedbydrift]

Define "healthcare" :-/

[Brant Peterson @fixedbydrift]

@mbeisen First came across the phrase "SES laundering" the other day in this context and can't stop worrying that's 90% of my models

[Brant Peterson @fixedbydrift]

Now if we could start making aluminum casts of crazy DNN architectures, at least a museum exhibit about twenty-first century computational science would have a nice aesthetic pitch

[Brant Peterson @fixedbydrift]

This is the analogy I've been reaching for all morning. The contribution of glass working to chemistry is FASCINATING (https://t.co/3HxoSGgG5D) but also approximately irrelevant to learning, doing, etc modern chemistry

[Michael 英泉 Eisen @mbeisen]

This is a bit like saying all science is, for all intents and purposes, a branch of plastic manufacturing. It’s technically true, and contains a germ of truth, but isn’t “true”.

[Sick Papes @SickPapes]

Types of #SICKPAPES - there’s nothing better

[Brant Peterson @fixedbydrift]

"some damn piece of research code that isn't done failing yet" is my new bio

[Chris Dwan (he / him) @fdmts]

Second: All queues need runtime limits. 48 hours is good. 24 is better.

If you waver on this, you'll eventually be unable to do maintenance because some damn piece of research code isn't done failing yet - after a month.

See above about long jobs blocking queues.

6/?

[Brant Peterson @fixedbydrift]

@Eric_Fauman I like this. It was always going to be messy to back the mechanistic effects of DNA variation into the epiphenomenal language of trait segregation genetics; at some point all these abstractions go from "leaky" to "sinking" :)

[Brant Peterson @fixedbydrift]

@Eric_Fauman A gene displaying divergent phenotypes over its alleles is then pretty clearly pleiotropic; the individual alleles less obviously so?

[Brant Peterson @fixedbydrift]

@Eric_Fauman I _think_ this is straightforwardly "alleles" (variants at a single gene locus).

Hesitant to call this "allelic series", since that implies phenotypic variation in magnitude but not in kind.

[Brant Peterson @fixedbydrift]

@mbeisen There's something awesomely revealing about "replace 'racism' with 'socioeconomic status'"

Like, state-a-valid-causal-model-of-US-history caliber revealing...

[Helen Czerski @helenczerski]

A friend has just shown me this book "Calculus made easy", published in 1914, and I think it's got one of the best prologues I've ever seen. This is *exactly* what textbooks should be doing. And they should all be honest about how terrifying the topic names are too.

[Brant Peterson @fixedbydrift]

My Twitter handle loves this result _almost_ as much as I do.

Also, for someone who grew up (in science at least) dreaming of a cis-reg M-K test, this 🧵 is like a message from a better future ✨

[Carl de Boer @CarldeBoerPhD]

Gene regulatory networks are highly interconnected; we asked how the TF input degree of promoters (regulatory complexity) changed under stabilizing selection on gene expression. Drift brings complexity extremes back to the average, arguing little selection on complexity itself.

[Brant Peterson @fixedbydrift]

Owl ETL

[Brant Peterson @fixedbydrift]

@PhilippJFR @HoloViews Ah 💯 `index_cols` in link_selections is what I was missing; I get the error this pr fixes, but this is definitely the behavior I was looking for thank you!

Now to figure out how to sort selected rows to the top of the table ;)

[Brant Peterson @fixedbydrift]

This can't be the right way to link selections between tables and plots in @HoloViews ? https://t.co/wyyCPWjri4

Must be some simpler notion of reciprocal shared selection, right? Like
tbl = hv.Table(...)
fig = hv.Scatter(...)
lsel = link_selections.instance()
lsel( fig + tbl )

[Brant Peterson @fixedbydrift]

@HoloViews paging @ericmjl :)