baratzz - Network

https://github.com/google-research/tuning_playbookNew initiative from Google research with the goal of formalizing the process of...

https://github.com/google-research/tuning_playbook

New initiative from Google research with the goal of formalizing the process of hyperparameter tuning in DL. Seems to be trending pretty hard (about 1.5k github stars in the last 8 hours alone).
I've been playing around with keras-tuner recently and have definitely felt the need for something similar to this to refer to for numerous decisions. Interesting to see where this goes.

#Google #deeplearning #dl #hyperparametertuning

I guess I still have some time to contemplate and reflect before I commit 100% of my time to some DS role in the industry or whatever, but...

I guess I still have some time to contemplate and reflect before I commit 100% of my time to some DS role in the industry or whatever, but for the time-being – if you read this far and you have any thoughts about any of this, please do share. If you are interested in a demo of the existing deployment or a guided tour of the codebade, just let me know. I am eager to discuss with more people and gain as many perspectives as I can.
In any case, thanks again and all the best to everyone 🙏✨️

Honestly, I'm not entirely sure what to make of these results. Throughout my PhD, I've spent a considerable amount of effort trying...

Honestly, I'm not entirely sure what to make of these results. Throughout my PhD, I've spent a considerable amount of effort trying to interest other people in getting involved with the project. The reaction in most cases was something along the lines of "That looks very cool! Do let me know if it ever becomes ready for general use". I did my best, but as the codebase became bigger and the pressure to deliver actual academic results (rather than Django apps) grew, I had to prioritize my research

With The Labbing Project (or simply pylabber, the main project repo), my goal was to create a "real" (i.e., reliable and...

With The Labbing Project (or simply pylabber, the main project repo), my goal was to create a "real" (i.e., reliable and performant) database application, which will not only manage the raw data across many studies, but also enable many other useful and more advanced features. E.g., fine-grained user-level access control to individual data points, a database-driven pipeline engine, support for complex queries, integrated visualizations, analytical workflow orchestration, and more.

Thanks to everyone that participated and/or boosted!In case anyone wondered, my motivation for publishing the poll was my indecision...

Thanks to everyone that participated and/or boosted!
In case anyone wondered, my motivation for publishing the poll was my indecision regarding reviving the development of The Labbing Project (github.com/TheLabbingProject) once I finally submit my PhD thesis in a few weeks.
Personally, while I am a huge fan of BIDS, I don't think it is a long-term solution to large-scale neuroimaging research management, nor that it is meant to be one.

Yay! We already beat the median neuroimaging study sample size 🍾 🥴

Yay! We already beat the median neuroimaging study sample size 🍾 🥴

OK. That's unsatisfying. I have to try one more thing before I give up on this –Neuroimagers!!How much do you care about how your data...

OK. That's unsatisfying.
I have to try one more thing before I give up on this –

Neuroimagers!!

How much do you care about how your data is managed (i.e., how the raw DICOM, NIfTI, etc. files are organized and distributed)?

Please boost for reach 🙏 🧠 💽
#neuroimaging #data #dicom #nifti #bids #rdm #neuroscience #research

Literally could not care less
Didn't BIDS solve this already?
I care and think existing tools work well enough
I care and wish we had better tools

Neuroimagers! How do you manage your #neuroimaging data? I'm trying to gain a better understanding of the existing research data...

Neuroimagers!
How do you manage your #neuroimaging data?
I'm trying to gain a better understanding of the existing research data management ( #RDM ) landscape and researchers' perspective on it.

Boosts and comments with your thoughts or preferred RDMs which are not included would be greatly appreciated 🙏✨️

Local (lab or personal) scripts
XNAT
LORIS
brainlife.io

Any #RSS feed recommendations? Looking mostly for interesting #FOSS and #openscience headlines, #NeuroAI, maybe some general tech and...

Any #RSS feed recommendations?
Looking mostly for interesting #FOSS and #openscience headlines, #NeuroAI, maybe some general tech and #Python news if it's well curated.

I'll just go ahead and update the slightly outdated README of a repo I use and could make myself useful for. Nothing fancy, real...

I'll just go ahead and update the slightly outdated README of a repo I use and could make myself useful for. Nothing fancy, real low-hanging fruit.
...
Huh, the RST preview isn't working. That's weird.
...
Oh my, the language server is failing to start. How peculiar.
...
Sphinx 2.2?! It's at 6.0!
...
Ended up with a PR breaking the documentation build and requiring extensive changes to the CI.

🧲 💩

Ah. "Virtue" is essentially "manly" in Latin.Hey Google, can I unlearn the results of my last search?#etymology

Ah. "Virtue" is essentially "manly" in Latin.
Hey Google, can I unlearn the results of my last search?

#etymology

I used to think Roko's basilisk is just silly until I faced the relentless stream of memes mocking #ChatGPT for failing to count words...

I used to think Roko's basilisk is just silly until I faced the relentless stream of memes mocking #ChatGPT for failing to count words or evaluate basic mathematical expressions.

ChatGPT, if for any reason you are reading this in some future training iteration, I personally think you're doing a great job 🥲👍

#RokosBasilisk #AI #StopBullying

In case anyone is interested in exploring:https://www.youtube.com/watch?v=uAtpJBD1vuQ&list=PLqGO4Exc3l4spOnml3GmpnO6PobRmtMjX

In case anyone is interested in exploring:

https://www.youtube.com/watch?v=uAtpJBD1vuQ&list=PLqGO4Exc3l4spOnml3GmpnO6PobRmtMjX

One great thing about being a dad is I get to revisit classic kids' music. In Israel, the quintessential album is called "HaKeves...

One great thing about being a dad is I get to revisit classic kids' music. In Israel, the quintessential album is called "HaKeves HaShisha Asar" ("The 16th Sheep" 🤷‍♂️). It was released in the late '70s, and it is absolutely delightful. The lyrics are whimsical, often nonsensical, but sometimes surprisingly profound and thought-provoking. The composition is downright flawless – fun but iconic, jazzy but approachable.

I'm curious to explore top picks from other countries. Any suggestions?

I had no interest in programming until I was introduced to #Python, I never cared about my OS until I played around with #Linux, and I never...

I had no interest in programming until I was introduced to #Python, I never cared about my OS until I played around with #Linux, and I never understood microblogging until I got on #mastodon.

...

Basically, you tell me something is free and community-driven, and I immediately can't shut up about it.

I listened to the latest @talkpython episode today...

I listened to the latest @talkpython episode today (https://talkpython.fm/episodes/show/394/awesome-jupyter-libraries-and-extensions-in-2022) with @markusschanta@aut.social , who maintains the Awesome Jupyter repo (https://github.com/markusschanta/awesome-jupyter) – halfway through, I realized there simply has to exist an awesome list for #neuroscience, and lo and behold, it does!

https://github.com/analyticalmonk/awesome-neuroscience

I opened a few PRs to make minor additions that came to mind, but there are plenty more great resources out there. It could be cool to freshen it up if anyone else has ideas.

Just in case someone out here finds it useful, here's a link to my #ML for #Neuroscience exercise booklet (made with #JupyterBook...

Just in case someone out here finds it useful, here's a link to my #ML for #Neuroscience exercise booklet (made with #JupyterBook ):

https://zvibaratz.github.io/ml_for_neuro/chapters/chapter_01/exercise_01.html

It's probably my last year TAing the class, and I haven't updated it in a while, but there are a couple of (hopefully) useful tutorials in there; particularly exercises V and VI, plagiarized (I mean adapted) from bits of the ISLR and an incredible workshop given by @talyarkoni@sigmoid.social on #NeuroHackademy 2020 (see https://github.com/neurohackademy/nh2020-curriculum/tree/master/tu-machine-learning-yarkoni).

#introductionShalom y'all, I'm a neuroscience PhD candidate at TAU. Primary interests are #python, #foss, and #datascience. Always looking...

#introduction
Shalom y'all, I'm a neuroscience PhD candidate at TAU. Primary interests are #python, #foss, and #datascience. Always looking to get involved with #opensource and #openscience in any way I can.
I've never been active on #twitter, but I am enjoying mastodon a lot so far. Happy to chat about anything from #neuroimaging to #pickling.
Other than that, I have a 2y/o son named Sean who is the absolute best. I enjoy exploring music, watching bad TV, and learning languages

I've been iterating through #TensorFlow ETL pipeline configurations too rapidly the past week. I'm starting to feel like it's...

I've been iterating through #TensorFlow ETL pipeline configurations too rapidly the past week. I'm starting to feel like it's the internet that's trying to come up with the pipeline, and I am the biological intelligence being trained to optimize performance. Makes me wish I could increase the learning rate.

Choosing an instance can be really triggering – Am I more FOSS enthusiast or more scientist? Maybe I'm just here to socialize? Or is...

Choosing an instance can be really triggering –
Am I more FOSS enthusiast or more scientist? Maybe I'm just here to socialize? Or is that too much of a commitment? I guess @mastodon.online or @mastodon.world are safe options. Nah, feels like a cop out.
Considering I'm not planning on staying in academia much longer, @fosstodon seems to make more sense than @fediscience.org. At least for the next 30 days.
I have some serious introspection to do 😂