
computational cognitive science @occidental
Joined February 2018
- Tweets131
- Following185
- Followers415
- Likes2.8K
My labs first preprint is out! Thanks @qywang71 and other great students
Does how you learn shape the geometry of what you know and your ability to update it?
180+ participants (EEG + Prolific), a few dozen RNNs, two phases: learn, then assemble.
biorxiv.org/content/10.64898…
1
4
158
Blocked learners built relational-certainty weighted geometries and more strongly expressed updated knowledge.
Interleaved learners paid behavioral costs, and those who successfully assembled showed high-D signatures and factorized representations
1
33
steph nelli retweeted
"Neural knowledge assembly in humans and neural networks"
@steph_nelli @summerfieldlab
Read more in @NeuroCellPress spkl.io/60144juWr
Our work concerning the learning and assembly of abstract knowledge is out (and much improved by reviewer comments) in @NeuroCellPress
new paper on how people (and neural networks) stitch ideas together. congratulations to lead authors @steph_nelli @ScienceLukas and collaborators @SaxeLab @atomsrivet cell.com/neuron/fulltext/S08…
also on (preprint)
#SfN2022 has begun. Interested in curriculum learning, contextual certainty or continual learning? Come check out my @Occidental students’ great work about how humans assemble knowledge tomorrow (Sunday) from 8-12
steph nelli retweeted
The under-attribution of women’s contributions to science is well known & well studied. A measure of under-attribution is the under-citation of papers authored by women relative to expected rates. In a new @NaturePhysics paper we explore this citation gap in contemporary physics.
It’s pretty great here- Apply!
1. Occidental is hiring an open-rank CS professor! oxy.edu/sites/default/files/…
2. I am going to Tapia for the first time next week! If you're interested in the position, or in working at liberal arts colleges, hit me up!
3
steph nelli retweeted
The Peters Lab is now open at Oxford and ready to build a team!
We'll study how the brain learns and controls behavior, especially in the cortex-basal ganglia-thalamus circuit.
peters-lab.org/
So proud of my students and the first (preliminary) results out of our lab @Occidental investigating how contextual certainty impacts knowledge assembly and continual learning 🎉
29
steph nelli retweeted
Our 2022 application cycle is now open for our #scholarships! If you, or a student you know, are #LGBTQ+ in a #STEM field, check out our different awards and their eligibility requirements. The portal closes June 4th. So apply today! https://nitter.cf/t.co/jisUJ7CO9k
steph nelli retweeted
The #learningSalon w @blamlab me @csuncodes is dlighted to host the inimitable @summerfieldlab
Friday 13th 4PM ET
Link: "What are we liiiike??"
crowdcast.io/e/learningsalon…
Recent work:
biorxiv.org/content/10.1101/…
biorxiv.org/content/10.1101/…
psyarxiv.com/qnpw6
Sure to be a great chat!
Excited (and nervous) to teach this upper level #deeplearning elective @Occidental this spring …
Here we go 🤩
2
28
steph nelli retweeted
Undergraduate summer internship listings, curated by @FUNfaculty!
funfaculty.org/undergrad_int…
steph nelli retweeted
Just out in the New Yorker: An article about Men Reading Minds with fMRI!
The author mentions 9 male and 0 female neuroscience professors, a feat made possible by impressive contortions like attributing a study from Ev Fedorenko’s lab to one of its male middle authors.
steph nelli retweeted
r u an undergrad interested in psychology and/or neuroscience and/or data science? r u URM &/or 1st-gen &/or low income? here's an idea for summer of 2022: psych.wisc.edu/nsf-reu-site-…
steph nelli retweeted
New preprint: Turns out neural networks are yet again pretty bad at something that humans are pretty good at: Reorganising neural representations based on a single new piece of information (Spoiler: We can fix it). Joint work with @steph_nelli @atomsrivet @SaxeLab @summerfieldlab
Neural knowledge assembly in humans and deep networks biorxiv.org/cgi/content/shor… #biorxiv_neursci
steph nelli retweeted
New preprint! With <2 mins training on two specific items, humans can reconfigure their neural reps of other items into a cohesive whole. Very cool. And very hard for deep networks.
Congrats @steph_nelli @ScienceLukas @atomsrivet & @summerfieldlab!



















