In my final year as a PhD student at this
#SfN23
, two of my PhD papers will be presented, i.e., SOTA tools to process and analyze brain-wide neural recordings.
These tools utilize convex optimization and support real-time interventions of thousands of neurons and BMIs. 1/5
My first PhD work is finally out at NeurIPS 2023, available here:
CORNN allows to train RNNs with tens of thousands of units, i.e., ~ 10-100 millions of parameters, to reproduce observed neural activity, roughly in real-time on laptops. 🧵incoming soon!
At first glance, this was clearly a joke. But apparently it is not!
I am selected an outstanding reviewer for Physica Scripta for 2020 and they will use my review to train potential reviewers 😳😳
Extremely proud of my childhood friend! I've personally witnessed his constant commitment and hard work, from age 5 to date, that led him to achieve remarkable success in life. It's a true testament to his dedication and perseverance. Your accomplishments inspire us all!
Regarding the protests at Bogazici University, Turkey; students share their plans for the future and their resentments from the unjust steps taken by the government. In solidarity with them, I too say:
"I no longer feel that I am listened to or wanted by my own country."
I am glad to share one of our recent works, now published in
@Nature
! Here, we studied the neural circuits of the placebo effect. I am happy to be part of such an amazing team, thank you Chong,
@GregScherrer25
, and Dr. Mark Schnitzer!
Time for a personal update!!!
Starting next year, I will be moving to UCSB as a
@KITP_UCSB
postdoc. I will be working with Dr. Boris Shraiman,
@ninamiolane
, and others I hope to meet along the way! Special thanks goes to
@baricandinc
for putting up with me during my PhD 🤗
Work co-supervised with
@Hidenori8Tanaka
, led by
@Udithhaputhant2
, has now been accepted to ICML Mechanistic Interpretability workshop.
TLDR: High-dimensional bifurcations underly skill acquisition in task-trained RNNs
Link:
A tweetprint🧵
Just received a desk reject (no interview) from a summer internship program on "Computational Neuroscience, Large Scale Simulation." 🥳🥳🥳
They didn't want me so much that they sent the rejection email twice...
After a long journey through 5 different countries, Beyza and I are finally settling down for the next few years at San Francisco.
We just finalized the contract, we found our dream home! Future is finally slightly less foggy 🐼
Stanford (neuroscience) has a website for undergraduates looking for research experience. PIs searching for students post their available projects here. Very helpful for remote/summer internships as well.
“Don’t work too hard, life is all about balance. If you plan ahead, you can do everything at once” type tweets are particularly getting to my nerves nowadays as I am failing to juggle both my PhD and social (?) life, while constantly struggling to keep both together...
Proud to be the husband of a superstar! Congratz to Beyza (
@baricandinc
) for receiving the highly competitive Jenessa Shapiro Graduate Research Award as a second year PhD and attracting a large attendance to her presentation!
Our lab at Stanford is on the hunt for a talented post-baccalaureate to join our team! This paid position is open to international applicants and requires a commitment of 2 years. No neuroscience background required, coding maturity is necessary and sufficient.
In 30 mins, at poster hall B
#NeurIPS2023
, I will talk about how we use RNNs trained instantly on neural activity from thousands of neurons to investigate the neural computation!
My first PhD work is finally out at NeurIPS 2023, available here:
CORNN allows to train RNNs with tens of thousands of units, i.e., ~ 10-100 millions of parameters, to reproduce observed neural activity, roughly in real-time on laptops. 🧵incoming soon!
I ran out of good things to say about neuroscience papers that I come across, so I will simply tweet them with a Pokemon name that I think suits the paper well. I will also keep them organized here, under this tweet by adding them as replies from time to time.
We are recruiting postdocs for 2024/25
@ai_ucsb
!
You want to build next-gen artificial scientific intelligence?🤖
Apply to the UCSB Real AI Initiative!
Deadline: Sept 15
Competitive salary
Prime location for your future office 🌊
@YaoQin_UCSB
@HaewonJeong00
@UofCalifornia
I just passed my quals and am finally a PhD candidate. During my quals, I was defending this paper:
While this was way out of my comfort zone, I am now up to date on superconducting qubits. It looks like the field has matured a lot, I am now a believer...
Today is THE day in Turkey. We celebrate our national sovereignty and it is a national holiday dedicated to children! To all those that are equally as excited about today:
#23NisanKutluOlsun
Everyone has their own PhD story. Mine was about becoming a family with
@baricandinc
and appreciation of shared interests with far superior others. I came to Stanford to become successful. Instead, I learned to enjoy life by doing what I love. Thank you Stanford 🙏
This is one of the few papers I am extremely jealous not to be an author of. Definitely an important read for anyone interested in learning dynamics and generalization in artificial (and biological) networks!
Congratulations
@RahulRam3sh
@EkdeepL
@KhonaMikail
@Hidenori8Tanaka
!
I’ll be at ICML in Vienna next week! Looking forward to presenting our work on compositional generalization in Transformers.
📑Arxiv:
💻Code:
🧵A thread summarizing our results.
If you know how to code in Python and have 2 hours to spare, this crash course is a very smooth first step into running basic ML algorithms in Python. Surprisingly, there is enough theory discussion to get someone started with limited background.
As my birthday gift, a group of my close friends decided to create a Wikipedia page for me. Best. Gift. Ever. 😂😂 (probably will be short-lived though, someone already tried to take it down)
Reading news from many talented young scientists about their PhD rejections is very disheartening. Being rejected from PhDs I applied to in my first try, I feel the pain. The pain is worse realizing that an application period costs more than twice the minimum wage in Turkey😟
Once again, instead of talking about things I have built a clear understanding over the last 2 years, I am going to discuss new concepts that I find extremely interesting nowadays.
I cannot promise that it will be good, but I can say that I am looking forward to it!
📢Next Tue at 10 am, we’ll have Fatih Dinç(
@fatihdin4en
) from Stanford University who will talk about the use of DL in understanding the neural circuitry in the brain:
"Uncovering The Neural Circuitry in The Brain Using Recurrent Neural Networks"
zoom info: ai-info
@ku
.edu.tr!
#AcademicChatter
#AcademicTwitter
Now is the time to fight for Turkish students’ rights! In a few days, this particular article and information within will be outdated, but human right violations occur everyday!
Thanks everyone for coming! I am around at
#NeurIPS2023
until saturday if you want to chat about RNNs and reverse engineering biological computations from large scale neural recordings!
📸 Yiqi Jiang
@extractneuro
Want to train RNNs on the large-scale, observed activity of tens of thousands of neurons in real-time? We're introducing CORNN! 🎉
Led by
@fatihdin4en
, A. Shai with M. Schnitzer from Stanford.
Paper link:
Meet us now at
#NeurIPS2023
: 10:45 AM, Tue, Dec
Honestly, this is how most of my friends feel as well and it has started way before the protests. We started our journey with the hope of going back home when the time is right. Now, most of us feels like there is no more "home" to go back to.
For this summer, we will be inviting three undergraduate interns to support our ongoing software development projects for processing large scale movies. Last year's project:
Details below
When I talk to undergrads regarding PhD or internship apps, I expect that the information will be disseminated further down to other students.
Today, I found out a student kept our conversion secret from close friends for comparative advantage. Don't do this, it is not worth it!
If you are interested in computational/theoretical aspects of neuroscience, come from a computational background with gaps in biological sides of things, then CNeuro is the summer school for you.
I attended last year, it has been very helpful and fun!
Grad school applications are around the corner and there is a plethora of opinions going around. Here is mine:
Read this before you apply. Most SOPs (especially many from Turkey that I had given feedback so far) suffer from the issues described here:
This study uses Extract, the cell extraction algorithm I helped developed!
I am so happy to see that my work saves people time and energy! Congratz to all involved and note that I am answering any questions related to Extract if you want to run it for your own data
#neurotwitter
Online now! Neurons switch from predominantly autonomous spiking to synchronous spiking in preparation for movement, a state change that emerges during skill learning
#neuroscience
This is my first contribution to neuroscience literature! I will be tweeting a thread about this paper very soon, explaining the main ideas behind the theory and providing a tutorial on the implementation. Stay tuned!
[TR] Ciddiyeti kenara bıraktığımız, iki eski dost geçmiş yıllardan çıkardığımız dersleri tartıştığımız çok ilginç bir yayın oldu. Yaklaşık bir 10dk kadar videoya girmemiş kahkaha görüntüleri mevcutmuş...
Üniversite tercihleri ile ilgili fermante üzüm suyu gibi yaşlanan görüşüm ektedir:
Bu dönemde insan hangi işi yaparsa yapsın iyi yapmalı. Neredeyse her alan interdisipliner hale geldi. Tercihler kadar sonrasını planlamaya odaklanmalı, çok da şey etmemek lazım…
If you haven't heard of the rising star
@mertyuksekgonul
, you will soon. I lost track of how many amazing papers he published in the last two years.
I cannot stress enough... Mert is at another level!!! I am proud to be a friend and envy his unprecedented level of resillience
Excited to share new
#iclr2023
Spotlight showing how to *surgically edit/debug*
#AI
w/o data or fine-tuning✨
We do this by turning standard DL ➡️ new class of model we call PCBM (no loss in perf). And it's easy to edit PCBM via high-level instruct🗣️
1/2
Reservoir computing: using chaos for computation
As the brain imaging technology matured and we looked carefully into activities at a single neuron level, we observed that finding ways to deal with chaos was inevitable for the brain.
Fig from: 1/7
I am fine mostly, but it really nerves me that PhD students are given advice either with the privileged mentality of 1980s or the ungrounded idealism of 2050.
If you are looking for a PhD in EE and/or CS, consider applying to UCSB! I cannot stress enough how kind, supportive, and incredibly smart everyone in Nina's group is!
Only a few days left to apply!
🧐How? Apply online to UCSB using one of the links given in the reply to this post (not via email).
🕑When? Deadlines:
- ECE, CS, Math, Physics: Dec 15th,
- DYNS (neuro): Dec 1st.
We look forward to your applications!😃
@UofCalifornia
In the 20th century, non-Euclidean geometry transformed how we model the world with pen and paper. In this century, it’s revolutionizing how we model the world with machines.
Our review on the topic is out , led w the brilliant
@naturecomputes
@johmathe
After having received a top 90th score from the GRE Math SUBJECT Test, I was told by a random prof to reconsider applying to Stanford because even high school students score higher and I should aim lower… Whether people care about GRE is a very robust red flag about them
People are taking the GRE math subject test soon.
If it goes poorly, just remember that it is an incredibly stupid test that has no predictive ability for how you’ll do at grad school research
A friend of mine went from a 5th percentile GRE score in 2008 to college prof in 2014
After taking this course last quarter (helped me read ML theory papers), I can safely suggest it for self-paced learning. The lecture notes are self-contained. This, however, is not an intro class and requires a substantial amount of time commitment
First time I was meeting with a prof. at Stanford, and I called someone genius. The prof. answered with:
“Einstein was a genius, the rest of us work hard and get the job done.”
My personal experience echoes this sentiment, not sure if being smart is as valuable as laymen think
Spending my whole day looking at awesome neural movies and trying to tune some hyperparameters and they said PhD is hard 😋
Jokes aside, looking at neurons fire and wondering why might be modern day’s equivalent of looking at the stars back in the 1600s
Bu yıl bana gelen EE mi CS mi sorularına cevabım: yazı tura atınız
Temel bilim okursam işsiz kalır mıyım: bu zihniyet üniversitede değişmezse mühendislikte de işiniz zor
TIP vs mühendislik: Büyük bir range’iniz var, her şeye beceriniz yetiyor demek ki. İkisinde de harcanırsınız
Why don’t we use complex numbers in machine learning research? Apparently, some people asked the same question and here is a relevant blog:
It seems that the biggest issue is the non-linearity, which cannot be both bounded, analytic and non-linear +++
3 stages of “knowing” I usually go through when I learn new concepts:
1) I understand what is written in this paper
2) I can explain it by writing a 5 page explanation
3) I can explain it with a single paragraph
It has taken more than a year in some cases to go from 2 to 3...
In undergrad, I was looking forward to days when I would no longer do homework.
In grad school, I am using my own free time to learn new topics and do self-assigned homework.
I toned down my future-oriented self during PhD to prevent similar situations, it works well so far!
New pre-print alert!
EXTRACT, a tractable and robust automated cell extraction tool for 1p and 2p calcium imaging!
#AcademicChatter
#neurotwitter
This also happens to be my first little contribution to neuroscience. A 🧵 1/6
(Video: Kim et al 2016)
[TR] Açık ofis saatini düzenli hale getireceğim, ancak şu sıralar taşınma işleri ile meşgulüm. Düzenli hale getirdiğimde de bunu temalı hale getirmeyi umuyorum, bazen beraber makale okuruz bazen kısa ders yaparım diye düşünüyorum. Ancak önce bir taşınabilmem lazım...
I have been burried deep in my own work for a while, thus my Twitter engagement has decreased down to checking new papers/retweeting them. I suspect this will be the case for a long time. On another note, my research has been very exciting lately, but I cannot share just yet!
Anıl ile uzun zamandır birlikte çalışıyoruz. Önümüzdeki yıl nörobilim alanında çok iyi bilinen bir grupta doktoraya başlıyor. Azmin ve doğru karakterin insanı çok ileriye götürebildiğine harika bir örnek. Başarılar Anıl, yolun açık olsun!
Lisans eğitimimin sonuna yaklaşırken sonraki durağım belli oldu! Eylül'de UC San Diego'da yapay öğrenme ve veri bilimi alanında bir doktora programına başlayacağım. 🎓 Lisans hayatım boyunca bana olan destek ve rehberlikleri için hocalarıma yürekten teşekkür ederim.
[TR] Bir saat oturup konuşalım derken 2 saat olmuş. Biraz bilim konuştuk biraz gelecek korkusundan bahsettik. Sanırım yakın zamanda bir ofis saati daha açacağım, çok keyifliydi!
I will be holding a public office hour in a few weeks. Feel free to join if you want to talk about some fundamental theoretical neuroscience topics and more!
Ben de 1-2 hafta içerisinde teorik sinir bilim üzerine tartışmak isteyenler için bir ofis saati açacağım.
I want to amplify two young scientists:
@AsafToprakc
and
@anilaslihak
grew amazingly in the last year we worked together but never met face to face. Both were first year students, a year later they are the co-authors of a manuscript we just submitted. Congratz to both👏👏
In today's discussion, I finally realized my position in the "linear algebra comes first vs coding comes first" debate on how to first learn ML. I think neither comes first, but critical thinking. Once one knows WHY they need ML, the rest comes quite easy no matter the path...
If you plan to apply for grad school upcoming fall, now is a good time to start thinking about which groups fit your interest the best.
Good chunk of research is done from home, so this summer is a great opportunity to send summer internship applications to those groups!
Academic excellence does not preceed rather usually proceeds academic freedom.
My alma mater has been under attack for more than two months now. But, here is the catch: If you push something, it pushes back equally as hard. Basic physics. Anyways, time to plant 🌲
Computational and Systems Neuroscience (Cosyne) conference will be online this year, it is starting Feb 23. There is also a tutorial on RNNs, so it should be fun. For more information and to register:
I have this habit of not being able to start preparing my talks until 24 hours to the presentation. It seems like an unsustainable practice, but somehow it always turns out ok...
I am listening to a grad admissions outreach in Turkey. It has been a whole hour of talks, and not one person has mentioned research fit and/or letter of recommendations. Cultural differences might be one of the biggest barriers for the grad school entry of internationals..
Statquest is very helpful for explaining the basic concepts of machine learning methods we frequently see in papers. It has a website and a Youtube channel dedicated to making ML concepts intuitively available to public and researchers.
I have posted my public office hour at
@acik_ofis_saati
, it will be on 23.00-23.59 GMT+3 (Turkey), 13.00-13.59 PT (California) this thursday, June 17th. Come and join for a very chaotic discussion ranging from neuroscience to physics to how I failed at making icecream at home🍦
🧠 Internship openings: NTT Research at Harvard! 🤖
Want to solve cutting-edge problems in deep learning by theory-guided algorithm design? Want to apply tools and ideas in ML to understand the brain?
Come join us this summer at the Center for Brain Science at Harvard!
1/
Q: Will we ever be able to predict exactly when and how strongly a single neuron fires?
This is a question that most young people, including me, entering the field of neuroscience wonders. Yet, when we look at the current models, we either see a) probabilistic b) chaotic models
If you plan to pursue a masters in AI related research, consider checking out the Vector institute. They have a scholarship program for AI masters and a list of AI master programs in Canada:
@AcademicChatter
#AcademicTwitter
I find this tutorial to be quick, to the point, and very accessible. If you want to transition into Python, this is a good starting point!
I list this as a study material for prereqs in my ML teachings
Python Tutorials for Absolute Beginners by CS Dojo
I keep finding myself revisiting this review and every time I learn something new:
Neural mechanisms of information storage in visual short-term memory
For this summer, we will be inviting three undergraduate interns to support our ongoing software development projects for processing large scale movies. Last year's project:
Details below
Interested in topological deep learning
#icml24
?
Join tomorrow's poster session to discuss the position paper:
🤖Topological Deep Learning is the New Frontier for Relational Learning 🍩
Led by
@theopapamarkou
with an amazing team of co-authors 🌟
Google translate works surprisingly well on this one.
So, if you are wondering why there are protests in Turkey regarding academic freedom and independence, read this carefully. Many of us feel the same way. An amazing 🧵by
@mervekahraman71
@lisachalik
Hey! I find Statquest to be very helpful in that regard:
Lately, it has been more about ML but there are some awesome videos on stats concepts as well!
A message from my long-time mentor and current collaborator.
The norms in the country changed so rapidly in the last 5-10 years that it makes no sense even for me, a recent graduate, to share my experience with current students, they are living in a different Turkey now…
@niloyka
@bernalacin35
Eğitim sistemimiz çoktan iflas etti. Yurtdışına gidiş artık lise mezuniyeti seviyesine indi.
Biz bile lisansüstü öğrenci bulmakta zorlanıyoruz. Kendi öğrencimiz yurtdışına gidiyor. Master için diğer üni.lerden bize gelişin nedeniyse Boğaziçi'nden yurtdışına kabul alabilmek.+++
The world didn’t leave much to celebration nowadays, so I will be quick. We have published a paper together with
@tuna_tugcu
@pusane
@fermat4214
and many more:
One of the easiest nicest thing you can do is to take part in research studies that use online surveys. It is literally few clicks and 10mins of your time, but you will make the day (or even the week) of the researcher running the experiment.
I think this is a very underrated class for non-engineering people. Some aspects might be too advanced, but the overall content is necessary for computational research
A primer on "Bayesian statistics and modelling" is just published on Nature. Looks like a must-read.
For many disciplines, primers provide a good intro to the hot topics of the field.
I guess when academics see "5 different countries" they will realize the pain, while others seeing this will be like "you lucky bastard!"
Well, I can say that it feels good to know that I will not need to move next year...
[TR] Ertunç beni bir dersine, hayatımdaki ilk seminerimi vermeye davet etmişti. Benim için bir ilkti ve çok heyecanlanmıştım, misafirperverliği ve harika karakteri sayesinde kendimi çok rahat hissetmiştim. Akademik kariyerinde harika yerlere geleceğine eminim, gelmeli de!
Sevgili Selen, mesajının benimle ilgili olan kısmı beni son derece duygulandırdı. Daha yolun çok başındayım. Akademik gelişimine en ufak bir katkım olduysa ne mutlu bana. Eğitim ve öğrenim hayatında başarılar diliyorum. Daha doğrusu başarılarının devamını diliyorum! Sevgiler.
Bu çarşamba da aynı saatte açtım. Bu dönemde doktora/yüksek lisans başvurusu yapacaklar için temayı genel başvurular olarak seçeyim dedim. Olası konular: Başvurulacak okullar nasıl seçilir, ilk mail nasıl atılır, CV/Sop nasıl hazırlanır, referans nasıl istenir vb.
I will be presenting the following short-talk in Kobit-5 this Friday. ()
While it has been some time I transitioned from quantum research to neuroscience, I am quite excited to share some recent observations on our 1y old work with
@agatabranczyk
My wife keeps telling I should start packing my luggage for the trip ahead. At first, I resisted her as best as I could. But later, thanks to her, I finally sorted all the papers in my to read file…
This is a fight for integrity, both academical (see plagiarism claims for the new rector) and social (see unlawful violations towards protesting rights). You can learn more up-to-date information from the alumni account:
I used to hate it when my old advisor picked up on the smallest details in figures. Today, as we are wrapping up a manuscript, my mentees feel the same...
That being said, details do matter and even if people do not care, we care. (Wow, I am becoming my old advisor...)
I was very busy lately and Beyza thought she would surprise me by cleaning the skillet and let it dry on its own. Oh well, work can wait, time for some CPR...
How to analyze comput. & dynamic mechanisms of RNNs?
1/2 Our
#NeurIPS2023
spotlight on a highly efficient algo for locating all fixed points, cycles, and bifurcation manifolds in RNNs:
@NeurIPSConf
By brilliant
@Zahra__Monfared
,
@lukasironman
, Nic Göring
Just a friendly reminder that the ability to follow the linear algebra behind ML algorithms does not automatically translate to a fundamental understanding.
Similarly, the simplicity of math behind an algorithm does not mean that it is trivial/requires little expertise. Ex 👇