Abdulkadir Canatar

Abdulkadir Canatar

Auto-generated profile, based on publicly available information

Abdulkadir Canatar

Flatiron Institute

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 May 2021 in Nature Communications
Verified
Journal Article
Open access
PDF file
ACAbdulkadir Canatar
Abdulkadir Canatar
BBBlake Bordelon
Blake Bordelon
CPCengiz Pehlevan
Cengiz Pehlevan
A theoretical understanding of generalization remains an open problem for many machine learning models, including deep networks where overparameterization leads to better performance, contradicting the conventional wisdom from classical statistics. Here, we investigate generalization error for kernel regression, which, besides being a popular machine learning method, also describes certain infinitely overparameterized neural networks. We use techniques from statistical mechanics to derive an analytical expression for generalization error applicable to any kernel and data distribution. We present applications of our theory to real and synthetic datasets, and for many kernels including those that arise from training deep networks in the infinite-width limit. We elucidate an inductive bias of kernel regression to explain data with simple functions, characterize whether a kernel is compatible with a learning task, and show that more data may impair generalization when noisy or not expressible by the kernel, leading to non-monotonic learning curves with possibly many peaks.

4 August 2025 in Physical Review E
Verified
Journal Article
Open access
ACAbdulkadir Canatar
Abdulkadir Canatar
SCSueYeon Chung
SueYeon Chung
A key problem in deep learning and computational neuroscience is relating the geometrical properties of neural representations to task performance. Here, we consider this problem for continuous decoding tasks where neural variability may affect task precision. Using methods from statistical mechanics, we study the average-case learning curves for ɛ -insensitive support vector regression and discuss its capacity as a measure of linear decodability. Our analysis reveals a phase transition in training error at a critical load, capturing the interplay between the tolerance parameter ɛ and neural variability. We uncover a double-descent phenomenon in the generalization error, showing that ɛ acts as a regularizer, both suppressing and shifting these peaks. Theoretical predictions are validated both with toy models and deep neural networks, extending the theory of support vector machines to continuous tasks with inherent neural variability.

24 February 2021 in Physical Review B
Verified
Journal Article
Closed access
PDF file
CYCenk Yanik
Cenk Yanik
VSVahid Sazgari
Vahid Sazgari
ACAbdulkadir Canatar
Abdulkadir Canatar
YVYaser Vaheb
Yaser Vaheb
IKİsMet İ. Kaya
İsMet İ. Kaya
A gate-induced insulating behavior at zero magnetic field is observed in a high-mobility suspended monolayer graphene near the charge neutrality point. The graphene device initially cleaned by a current annealing technique underwent a thermopressure cycle to allow short-range impurities to be adsorbed directly by the ultraclean graphene surface. The adsorption process generated a strong temperature- and electric-field-dependent behavior on the conductance of the graphene device. The conductance around the neutrality point is observed to be reduced from around e2/h at 30 K to ∼0.01e2/h at 20 mK. A direct transition from an insulator to a quantum Hall conductor within ≈0.4T accompanied by broken-symmetry-induced ν=0,±1 plateaus confirms the presence of intervalley scatterers.

1 January 2025
Verified
Conference Paper
Closed access
CCChanwoo Chun
Chanwoo Chun
ACAbdulkadir Canatar
Abdulkadir Canatar
SCSueYeon Chung
SueYeon Chung
DLDaniel Lee
Daniel Lee