Philipp Berens
Professor of Data Science @ University of Tübingen, Director of Hertie AI (www.hertie.ai) and Speaker of ML4Science (www.machinelearningforscience.de)
- [Not loaded yet]
- [Not loaded yet]
- We did just that a few times - successfully sometimes, unsuccessfully some other times
- The University of Sydney is recruiting senior and mid-career academics "within the top 5–10% of academics in their field, with a strong track record and availability to relocate to Australia within 6-months". If you're a psychologist/neuroscientist and are interested, get in touch with me soon.
- how do you provide evidence you are in the top 5-10% of your field?
- [Not loaded yet]
- [Not loaded yet]
- You were forced to review with >4 papers - with rejection of all of you didn’t comply. But again, the tone was rough.
- bsky.app/profile/phil... I keep getting mails of my co-authors asking me what I did - on the other hand, the ACs response was kind. I was flagged automatically apparently… harsh for public shaming
- I was flagged as "irresponsible" reviewer at NeurIPS for a paper providing no algorithm, no theory, no data, no analysis... it is simply an opinionated, colloquial review. I said as much. Instead of looking at it, the AC forced me to add to my review, threatening to reject papers of my co-authors.
-
View full thread🙃🙃🙃🙃
- And don't get me wrong: I think it is good that they are trying to improve review quality.
- What??? That sounds absurd
- And they literally send it to ALL MY COAUTHORS, telling them I submitted an irresponsible review. I mean...
- [Not loaded yet]
- 🇩🇪🇨🇭🇺🇸🇪🇸🇳🇬🇨🇲🇬🇭🇷🇺🇮🇳🇨🇳🇨🇴🇨🇦
- [Not loaded yet]
- [Not loaded yet]
- Until Trump outlaws the practice of eating vegan or vegetarian
- [Not loaded yet]
- The problem is that you will only vote again in 3.5 years - if you vote again
- [Not loaded yet]
- Oh boy
- @ml4science.bsky.social is currently 17th in the 100k relay race of @unituebingen.bsky.social! 14 laps to go! #ai4science
-
View full threadStunning 18th place in finish! @ml4science.bsky.social
- The colleagues from the new excellence cluster TERRA @unituebingen.bsky.social arriving after 100k on the 9th place!
- [Not loaded yet]
- Oh it is actually very nice, you can just prompt it to use Newspeak
- [Not loaded yet]
- [Not loaded yet]
- Yeah, that’s an issue that you cannot really rely on it. But then this is true for so many things with DB. At a Kopfbahnhof like Frankfurt or Leipzig (?) you may be able to grab something from the station
- [Not loaded yet]
- Really? I like the vegan pasta and curry options!
- [Not loaded yet]
- [Not loaded yet]
- Depending on your perspective 80y is a long time
- [Not loaded yet]
- Just because you criticize neoliberalism you are not automatically Marxist
- [Not loaded yet]
- 80y later or so…. Up to the late 1920ies, Germany was the go to nation worldwide for maths, philosophy & even biologists published in Zeitschrift f Naturforschung… up until very recently a German academic career needed at least 2y in the US
- [Not loaded yet]
- [Not loaded yet]
- Now you are laughing, but who knows.
- When I went to the US in 1998 as a high school student, the "land of the free" offered remarkable new perspectives on my home country Germany. As a liberal-left German, the flags, the pledge of allegiance and patriotism were strange, yet also fascinating in a weird way.
- Now, the US government is halting student visa scheduling, increases social media surveillance, and patriotism has turned into cheap, blatent nationalism and facism following a great leader or manipulating him for ones own gain.
- For students in my kids' generation, I don't see how going on an exchange to the US holds much appeal. We need to be vigilant to similar tendencies in Germany and Europe, and stop the far-right from gaining further ground.
- There should be a categorical imperative for bikes: Treat your bike like you would want to be treated by it
- [Not loaded yet]
- [Not loaded yet]
- Can you send me an email? We just got an extension to uni-tuebingen.de/forschung/fo...
- We are incredible happy to be able to continue our work of developing new #AI4science across a wide range of disciplines with incredible colleagues in #physics, #neuroscience, #cogsci, #geoscience, #linguistics, #economics, #medicine, #philosophy, #law and #anthropology! @unituebingen.bsky.social
- We're super happy: Our Cluster of Excellence will continue to receive funding from the German Research Foundation @dfg.de ! Here’s to 7 more years of exciting research at the intersection of #machinelearning and science! Find out more: uni-tuebingen.de/en/research/... #ExcellenceStrategy
- All NSF GRFPs to students choosing to attend Harvard in the Fall were just terminated
- [Not loaded yet]
- Put Tübingen on the list as well (close by American standard to Heidelberg) and a nice Uni
- We all need to watch carefully what is happening in the US and what could happen if our own fascists come into power (again)
- And make no mistake about it... Attacking higher education like this, particularly a private institution to try and force them to bend a knee to your will is verrrrry fascist. en.wikipedia.org/wiki/Fascism
- This is where @hertie-ai.bsky.social will be located in three years or so!
- And another one: We show how to use a diffusion model for #fundus images for #counterfactual reasoning in #ophthalmology! If you ever wondered how the fundus images of your patient had looked, if he had been suffering from diabetic retinopathy, check this out ⬇️ journals.plos.org/digitalhealt...
-
View full threadJoint work with PhD students Indu Ilanchezian & Valentin Boreiko, as well as Ziwei Huang @msayhan.bsky.social @lisakoch.bsky.social Matthias Hein and Laura Kühlewein!
- PS. @plos.org #DigitalHealth is a great journal for this kind of work run by a community of dedicated domain experts with diverse backgrounds!
- This allows to generate high quality counterfactual images for individual patients, e.g. answering the question how an image had looked if the patient had been healthy or if she had been sick. The induced lesions look highly realistic, and the vessel structure is kept intact.
- Importantly, we do not simply compute quality metrics, but perform a highly sensitive three-way odd-one-out study to assess the realism - even for ophthalmologists, it is hard to figure out, which images have been generated by our algorithm!
- In our model, we use the gradients of a robust disease classifier and regularization towards the original image to guide the diffusion process
- [Not loaded yet]
- [Not loaded yet]
- Every Coding Lab has a section where they are supposed to report what they used, how and for what. It seems that the reporting works quite well. Send me an email and I can share prelim summary (by LLM of course).
- [Not loaded yet]
- CELL Press has multi-journal submission. Works actually very well. Papers are reviewed at up to three journals, and you get editorial feedback from each.
- We told the students to use them however they like and just tell us how - very interesting results! Grades are about the same as before…
- [Not loaded yet]
- And it’s so great they finally put the moose 🫎 front and center as a model animal for retina research! It’s just so handy!
- We had a paper accepted today! In the first round, Rx provided a long, forceful rant why our paper wasn't a big advance, how the figures are poorly done and so on. We pushed back politely and answered the points we thought were reasonable. Today, Rx: "Sorry for my rant, the paper is good." Cudos!
- Thanks Christian!
- Putting this on top of my reading list: Task-specific regional circuit adaptations in distinct mouse retinal ganglion cells | Science Advances www.science.org/doi/10.1126/...
- We are excited that our work on inherently interpretable #deeplearning models for #AI for #medicine has been published in @plos.org #digitalhealth! You want to know how to combine the power of deep learning with accessible interpretation? This is for you! ⬇️ journals.plos.org/digitalhealt...
-
View full threadThis not only works much better than established post-hoc techniques, but also works well as a support tool for clinicians: Decisions for difficult cases improve and all decisions become about 20% faster!
- Thus this work shows how #inherently #interpretable models in #AI for #medicine can improve clinical decision making in #ophthalmology! Thanks to everyone involved, especially Kerol Djoumessi (hertie.ai/data-science...) and @lisakoch.bsky.social!
- BagNets are modified ResNets with local receptive fields and an explicit class evidence maps at the end. We discussed there use for interpretable #medical #AI here: proceedings.mlr.press/v227/donteu2...
- The explicit class evidence maps allows to penalize activations, e.g. via a sparseness penalty. This is extremely effective for diseases like #diabetic #retinopathy where lesions are small in initial disease stages: extracted high evidence regions almost always contain lesions.
- Sigh...
