Article
Evolutionary Biology
Joseph Waterton, Susan J. Mazer, Justin R. Meyer, Elsa E. Cleland
Summary: The text discusses the trade-offs between fitness-related traits and how certain trait combinations can constrain adaptive evolution, especially in more arid environments. Through a greenhouse experiment with two grass species, the researchers found a trade-off between within-year emergence speed and potential among-year emergence spread, highlighting the importance of early emergence and larger seed persistence fraction for adaptive responses in arid environments. This study demonstrates how genetic correlations within populations and the concept of Pareto optimality can be used to detect evolutionary constraints and adaptation across environmental gradients.
EVOLUTIONARY APPLICATIONS
(2021)
Article
Behavioral Sciences
Anthony M. Jakob, John G. Mikhael, Allison E. Hamilos, John A. Assad, Samuel J. Gershman
Summary: The role of dopamine as a reward prediction error signal in reinforcement learning tasks has been well-established, and it also affects the speed of subjective time. According to the theory, the timing of dopamine relative to reward delivery determines whether subjective time speeds up or slows down. Reanalyzing measurements of dopaminergic neurons in mice performing a self-timed movement task, it was found that dopamine activity timing could predict changes in subjective time speed.
BEHAVIORAL NEUROSCIENCE
(2022)
Article
Neurosciences
Ezgi Gur, Alihan Erdagi, Fuat Balci
Summary: Many timing tasks favor high temporal accuracy but lack asymmetric cost functions. Asymmetric cost functions can result in adaptive biases in timed responses due to timing uncertainty. Differential reinforcement of response duration (DRRD) is a task that requires mice to actively respond for a minimum amount of time, and the bias in response durations is predicted by the level of timing uncertainty. Our study tested mice in a DRRD task and found that response durations were positively biased and correlated with endogenous timing uncertainty. These results contribute to the understanding of optimal timing behavior in non-human animals.
TIMING & TIME PERCEPTION
(2023)
Article
Computer Science, Interdisciplinary Applications
Francesca Cenci, Arun Pankajakshan, Pierantonio Facco, Federico Galvanin
Summary: The trade-off between experimental design space exploration and information maximization is still an open question in optimal experimental design. In this study, we propose a novel model-based design of experiments method that enhances space exploration and reduces model prediction uncertainty by using a mapping of model prediction variance (G-optimality mapping). The results show that this method outperforms classical design of experiments methods in terms of model prediction uncertainty reduction and parameters precision maximization.
COMPUTERS & CHEMICAL ENGINEERING
(2023)
Article
Biochemical Research Methods
John G. Mikhael, Lucy Lai, Samuel J. Gershman
Summary: This study introduces the concept of 'rational inattention' to the dopamine literature, which reconciles two influential theories of tonic DA under a unified framework. The rational inattention framework can explain a wide range of experimental findings, including reinforcement learning and interval timing effects, and provides a better understanding of how DA influences behavior.
PLOS COMPUTATIONAL BIOLOGY
(2021)
Article
Neurosciences
Rehan B. Chinoy, Ashita Tanwar, Dean V. Buonomano
Summary: Interval discrimination is essential in sensory processing, such as speech and music. Previous models suggest separate circuits for timing and working memory components. However, this study shows that the same recurrent neural network can implement both components.
TIMING & TIME PERCEPTION
(2023)
Article
Psychology, Biological
Dobromir Rahnev, Kobe Desender, Alan L. F. Lee, William T. Adler, David Aguilar-Lleyda, Basak Akdogan, Polina Arbuzova, Lauren Y. Atlas, Fuat Balci, Ji Won Bang, Indrit Begue, Damian P. Birney, Timothy F. Brady, Joshua Calder-Travis, Andrey Chetverikov, Torin K. Clark, Karen Davranche, Rachel N. Denison, Troy C. Dildine, Kit S. Double, Yalcin A. Duyan, Nathan Faivre, Kaitlyn Fallow, Elisa Filevich, Thibault Gajdos, Regan M. Gallagher, Vincent de Gardelle, Sabina Gherman, Nadia Haddara, Marine Hainguerlot, Tzu-Yu Hsu, Xiao Hu, Inaki Iturrate, Matt Jaquiery, Justin Kantner, Marcin Koculak, Mahiko Konishi, Christina Koss, Peter D. Kvam, Sze Chai Kwok, Mael Lebreton, Karolina M. Lempert, Chien Ming Lo, Liang Luo, Brian Maniscalco, Antonio Martin, Sebastien Massoni, Julian Matthews, Audrey Mazancieux, Daniel M. Merfeld, Denis O'Hora, Eleanor R. Palser, Boryslaw Paulewicz, Michael Pereira, Caroline Peters, Marios G. Philiastides, Gerit Pfuhl, Fernanda Prieto, Manuel Rausch, Samuel Recht, Gabriel Reyes, Marion Rouault, Jerome Sackur, Saeedeh Sadeghi, Jason Samaha, Tricia X. F. Seow, Medha Shekhar, Maxine T. Sherman, Marta Siedlecka, Zuzanna Skora, Chen Song, David Soto, Sai Sun, Jeroen J. A. van Boxtel, Shuo Wang, Christoph T. Weidemann, Gabriel Weindel, Michal Wierzchon, Xinming Xu, Qun Ye, Jiwon Yeon, Futing Zou, Ariel Zylberberg
NATURE HUMAN BEHAVIOUR
(2020)
Review
Neurosciences
Andrew Saxe, Stephanie Nelli, Christopher Summerfield
Summary: Deep neural networks offer potential theories for perception, cognition, and action in biological brains, potentially reshaping our understanding of neural systems. This Perspective provides a roadmap for how neuroscientists can utilize deep networks to model and understand biological brains.
NATURE REVIEWS NEUROSCIENCE
(2021)
Article
Cell Biology
Bilgehan Cavdaroglu, Sadia Riaz, Yuqing Shi, Fuat Balci, Rutsuko Ito
Summary: The study found that the ventral hippocampus CA3 subregion plays a critical role in regulating timing uncertainty in temporal decision-making by reducing decision threshold variability. This research suggests that vCA3 may be important in modulating decision threshold or switch closure latency variability.
Article
Neurosciences
Timo Flesch, Keno Juechems, Tsvetomira Dumbalska, Andrew Saxe, Christopher Summerfield
Summary: This study investigates how neural populations code for multiple conflicting tasks and proposes lazy and rich coding solutions. The findings suggest that the rich learning regime, which prioritizes relevant features, is consistent with neural coding patterns in biological brains.
Editorial Material
Multidisciplinary Sciences
Fuat Balci
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2022)
Article
Behavioral Sciences
Pinar Toptas, Ezgi Gur, Fuat Balci
Summary: Numerical and temporal control of behavior is widespread in animals. Mice ignore temporal relations and probabilistic information in the presence of reliable discriminative stimuli, but start relying on previously experienced time intervals and probabilities when discriminative stimuli become non-informative. However, similar dynamics do not apply to counting behavior due to differences in reinforced outcomes based on the number vs. timing of responding. This study found that even in conditions with strong stimulus control, mice showed a relatively strong representational control over their counting behavior.
Article
Psychology
Ece Yallak, Fuat Balci
Summary: Recent research suggests that humans can track the direction and magnitude of their timing errors without feedback. This study provides indirect evidence of temporal error awareness and a strong incentive to maximize timing accuracy. The findings show that participants opted out of trials with larger distances between their reproductions and target time intervals, resulting in lower timing precision.
JOURNAL OF EXPERIMENTAL PSYCHOLOGY-HUMAN PERCEPTION AND PERFORMANCE
(2022)
Article
Behavioral Sciences
Ezgi Gur, Yalcin A. Duyan, Fuat Balci
Summary: This study investigated whether mice can make temporal inferences of novel locations based on previously learned spatiotemporal contingencies. The results showed that mice exhibited a response pattern at new locations that fell between the time intervals of previously reinforced locations.
Article
Biology
Javier Masis, Travis Chapman, Juliana Y. Rhee, David D. Cox, Andrew M. Saxe
Summary: Balancing short-term speed and accuracy is crucial for making optimal decisions in the presence of noise. This study demonstrates the importance of long-term learning in the speed-accuracy trade-off and provides a theoretical framework that incorporates learning dynamics. The findings reveal that choosing suboptimal response times to facilitate faster learning can lead to greater total reward, suggesting cognitive control over the learning process.
Article
Neurosciences
Weinan Sun, Madhu Advani, Nelson Spruston, Andrew Saxe, James E. Fitzgerald
Summary: Memorization and generalization are cognitive processes that promote adaptive behavior. Animals memorize safe routes to water sources and generalize from these memories to predict new ones. Systems consolidation mechanisms construct neocortical memory traces, but why it only applies to a subset of memories is unclear. This study proposes that memories only consolidate when it aids generalization, providing new insights into adaptive behavior.
NATURE NEUROSCIENCE
(2023)
Article
Psychology, Educational
Fuat Balci, Gokce Elif Baykal, Tilbe Goksun, Yasemin Kisbu, Asim Evren Yantac
Summary: Many training programs aim to improve creative thinking abilities, but there are fewer programs for children compared to adults. This study implemented a nine-week long creativity intervention program for socioeconomically disadvantaged children in Turkey, and the results showed that children in the experimental group had significantly higher verbal fluency and originality scores compared to the control group.
CREATIVITY RESEARCH JOURNAL
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Andrew M. Saxe, Shagun Sodhani, Sam Lewallen
Summary: In this work, the authors introduce the Gated Deep Linear Network framework to investigate the impact of network architecture on learning dynamics. They derive exact solutions to the dynamics of learning and demonstrate that structured networks can be conceptualized as a neural race with a bias towards shared representations, affecting the model's ability to generalize, multi-task, and transfer. This research provides insights into the relationship between neural architecture and learning, as well as the design of more complex architectures.
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162
(2022)
Article
Computer Science, Artificial Intelligence
Federica Gerace, Luca Saglietti, Stefano Sarao Mannelli, Andrew Saxe, Lenka Zdeborova
Summary: The study proposes a solvable synthetic data model to explore the correlation between transfer learning and generalization performance, and investigates the impact of feature transfer on generalization under different conditions.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)
Article
Behavioral Sciences
Ezgi Gur, Yalcin Akin Duyan, Fuat Balci
Summary: The study found that mice can average independently trained numerosities, especially when faced with conflicting information, by counting the number of required responses. The behavior of the majority of the mice in the test trials was explained by a counting strategy rather than a timing strategy. The number of responses in the test trials was equally well accounted for by arithmetic, geometric, and Bayesian averages.
Article
Psychology, Experimental
Tutku Oztel, Terry Eskenazi, Fuat Balci
Summary: A key aspect of metacognition is the ability to monitor performance, with recent studies focusing on error-monitoring ability to capture timing errors. Research used a more direct measure for temporal error monitoring (TEM) and tested the impact of feeling watched on performance, but no significant influence of social stimulus was found. In conclusion, metric error monitoring is a robust metacognitive phenomenon not easily influenced by social factors.
PSYCHOLOGICAL RESEARCH-PSYCHOLOGISCHE FORSCHUNG
(2021)