Mick Bonner
@michaelfbonner
Followers
902
Following
298
Media
32
Statuses
240
Assistant Professor @jhucogsci studying human vision using cognitive neuroscience and machine learning
Baltimore, MD
Joined January 2019
Can we gain a deep understanding of neural representations through dimensionality reduction? Our new work shows that the visual representations of the human brain need to be understood in high dimensions. w/ @RajThrowaway42 & Brice Ménard.
3
35
117
RT @TimKietzmann: Exciting new preprint from the lab: “Adopting a human developmental visual diet yields robust, shape-based AI vision”. A….
0
47
0
Excited to announce that this paper is now out in @ScienceAdvances
What do artificial and biological vision have in common? Shared architectures? Shared task objectives? In this preprint, we suggest a more general explanation: they share universal dimensions of natural image representation.
1
8
45
RT @martin_hebart: Very happy to announce that our paper comparing dimensions in human and DNN representations is now out in @NatMachIntell….
0
21
0
RT @martin_hebart: I'm thrilled to see this preprint out! Lenny compellingly demonstrates, in a data-driven way, a coding scheme unifying d….
0
5
0
RT @nmblauch: What shapes the topography of high-level visual cortex?. Excited to share a new pre-print addressing this question with conne….
0
30
0
RT @lu_zejin: Now out in Nature Human Behaviour @NatureHumBehav: “End-to-end topographic networks as models of cortical map formation and h….
0
15
0
Exciting new preprint from the lab, led by an amazing student, @ckelseyhan!.
How do different brains create unique visual experiences from identical sensory input? New preprint w/ @MichaelFBonner: "High-dimensional structure underlying individual differences in naturalistic visual experience" 🧠
0
3
9
RT @martin_hebart: Our lab is present at #VSS2025, so let me highlight some of the things we have been up to with a talk and three posters.….
0
1
0
RT @StphTphsn1: Better artificial intelligence does not mean better models of biology.Drew Linsley, Pinyuan Feng, Thomas Serre. https://t.co….
0
18
0
RT @NeuroKathyG: Excited to be presenting this paper at #ICLR2025 this week @iclr_conf! Come check our poster if you want to know more!! 🧠🤖….
0
4
0
RT @amarvi_: **ecstatic** to share our @iclr_conf paper: sparse components distinguish visual pathways & their alignment to neural networks….
0
7
0
RT @sparuniisc: In a study now out in @eLife, @GeorginJacob @PramodRT9 and I have some exciting results: a novel computation that helps the….
0
19
0
RT @martin_hebart: People talk a lot about objects, but what about the softness of a cushion, the greenness of an emerald, or the viscosity….
0
20
0
RT @sandervanbree: What does it mean when neural network dimensions converge? In this blog post I explore possible….
0
5
0
4/4 The ED that is measured in the Elmoznino & Bonner paper is the ED of channels, with spatial dimensionality factored out. This tells us about the richness of the features that a model learned, not simply the dimensionality of the regressors in an encoding model.
0
0
8
3/4 One needs to be careful when comparing ED across diverse models. There are model-specific factors that influence ED. When examining the relationship between ED and other factors of interest, you will get a much clearer picture if you control for architecture.
1
0
3
See our supplement section titled “High ED alone is not sufficient to yield strong performance.”.
1
0
3