Research
My research interests lie at the intersection of experimental and computational neuroscience. I aim to reveal a mechanistic understanding of the network algorithms used by both artificial and biological neural networks — studying how populations of neurons encode information, how learning shapes neural dynamics, and how computational principles are shared across artificial and biological systems.
Publications & Preprints
Learning rate collapse prevents training recurrent neural networks at scale
B. Kurtkaya, M. Harmanli*, A. Cimen*, A. Alexander, N. Miolane, F. Dinc†, Y. Yemez†
NeurIPS'25 – NeurReps Workshop · 2025
* equal contribution · † equal supervision
Conference Presentations
Dimensionality of population-level latent mechanisms encoding spatial representations
N. E. Delikkaya*, A. Cimen*, F. Acosta, A. Myers, A. Alexander, F. Dinc†, N. Miolane†
NeurReps'25 Conference · Poster · 2025
NeuroZoo: A tool for training recurrent neural network models on behavioral tasks
A. Cimen, M. Yuksekgonul, A. S. Alexander†, F. Dinc†
Society for Neuroscience (SFN) · Poster · 2025
Zoo of RNNs: A comprehensive analysis of recurrent neural networks trained on synthetic behavioral tasks
H. Akengin*, A. Cimen*, M. Yuksekgonul, A. Alexander, F. Dinc†
Society for Neuroscience (SFN) · Poster · 2024
* equal contribution · † equal supervision
Research Experience
Graduate
Graduate Researcher, The Alexander Lab
Sept 2026 – PresentUniversity of California, Santa Barbara
PhD research under Dr. Andy Alexander.
Undergraduate
Undergraduate Researcher, Geometric Intelligence Lab
June 2025 – Feb 2026University of California, Santa Barbara
Under Dr. Nina Miolane and Dr. Fatih Dinc. Studied why training large recurrent neural networks (RNNs) remains challenging at scale. Identified key sources of learning-rate collapse and demonstrated how population-coding frameworks mitigate these limitations. Work published at NeurReps 2025.
Undergraduate Researcher, Medical and Biological Physics Lab
Dec 2024 – May 2026Ozyegin University
Under Dr. M. Burcin Unlu. Investigated population dynamics that emerge during learning in the brain using context-dependent silico-tasks and recurrent neural networks, applying mathematical tools to uncover latent dynamics.
Undergraduate Researcher, The Alexander Lab
July 2024 – Feb 2026University of California, Santa Barbara
Under Dr. Andy Alexander. Received hands-on training in experimental neuroscience including mouse handling, behavioral training, cranial-window preparations, Mini2P calcium imaging, and electrophysiology. Designed an experiment to study how mice perform path integration on a circular periodic track. Trained RNNs on synthetic path-integration tasks and found they spontaneously learn Fourier-like basis functions. Preliminary results presented at NeurReps'25.
Undergraduate Researcher, Multirobot Intelligence & Perception Lab
Feb 2024 – Aug 2024Ozyegin University
Under Dr. Sedat Ozer and Dr. Fatih Dinc. Developed a comprehensive benchmark of behavioral tasks for RNN models and standardized traditional computational neuroscience tasks. Collaborated with teams from Ozyegin University, Stanford University, and UC Santa Barbara. Presented at SFN'24.