A5087212561- Neural dynamics and brain function
- Neural and Behavioral Psychology Studies
- Olfactory and Sensory Function Studies
- Decision-Making and Behavioral Economics
- Functional Brain Connectivity Studies
- Experimental Behavioral Economics Studies
- Memory and Neural Mechanisms
- Visual perception and processing mechanisms
- Advanced Chemical Sensor Technologies
- EEG and Brain-Computer Interfaces
- stochastic dynamics and bifurcation
- Advanced Memory and Neural Computing
- Behavioral and Psychological Studies
- Neuroscience and Neuropharmacology Research
- Neurobiology and Insect Physiology Research
- Receptor Mechanisms and Signaling
- Neural Networks and Applications
- Text and Document Classification Technologies
- Plant and Biological Electrophysiology Studies
- Stress Responses and Cortisol
- Insect Pheromone Research and Control
- Speech Recognition and Synthesis
- Neuroscience and Music Perception
- Face Recognition and Perception
- Game Theory and Applications
McGill University
2024-2025
Harvard University
2018-2025
Mila - Quebec Artificial Intelligence Institute
2025
Center for Pain and the Brain
2020-2023
Cold Spring Harbor Laboratory
2014-2022
Harvard University Press
2022
A.S. Watson (Netherlands)
2020
Learning from successes and failures often improves the quality of subsequent decisions. Past outcomes, however, should not influence purely perceptual decisions after task acquisition is complete since these are designed so that only sensory evidence determines correct choice. Yet, numerous studies report outcomes can bias decisions, causing spurious changes in choice behavior without improving accuracy. Here we show effects reward on principled: past rewards future choices specifically...
Odor cues in nature are sparse and highly fluctuating due to turbulent transport. To investigate how animals perceive these intermittent cues, we developed a behavioral task which head-restrained mice made binary decisions based on the total number of discrete odor pulses presented stochastically over several seconds. Mice readily learned this task, their performance was well-described by widely used decision models. Logistic regression choices against timing within respiratory cycle...
To thrive in complex environments, animals and artificial agents must learn to act adaptively maximize fitness rewards. Such adaptive behavior can be learned through reinforcement learning
Uncertainty is a fundamental aspect of the natural environment, requiring brain to infer and integrate noisy signals guide behavior effectively. Sampling-based inference has been proposed as mechanism for dealing with uncertainty, particularly in early sensory processing. However, it unclear how reconcile sampling-based methods operational principles higher-order areas, such attractor dynamics persistent neural representations. In this study, we present spiking network model head-direction...
In order to respond reliably specific features of their environment, sensory neurons need integrate multiple incoming noisy signals. Crucially, they also compete for the interpretation those signals with other representing similar features. The form that this competition should take depends critically on noise corrupting these study we show type commonly observed in systems, whose variance scales mean signal, selectively divide input by predictions, suppressing ambiguous cues while...
Torben Ott1,4, Paul Masset1,2,3,4 and Adam Kepecs1 1Cold Spring Harbor Laboratory, Cold Harbor, New York 11724, USA 2Watson School of Biological Sciences, 3Department Molecular Cellular Biology & Center for Brain Science, Harvard University, Cambridge, Massachusetts 02138, Correspondence: kepecs{at}cshl.edu ↵4 These two authors contributed equally.
Rational decision makers aim to maximize their gains, but humans and other animals often fail do so, exhibiting biases distortions in choice behavior. In a recent study of economic decisions, humans, mice, rats were reported succumb the sunk cost fallacy, making decisions based on irrecoverable past investments detriment expected future returns. We challenge this interpretation because it is subject statistical form attrition bias, observed behavior can be explained without invoking...
Abstract Within a single sniff, the mammalian olfactory system can decode identity and concentration of odorants wafted on turbulent plumes air. Yet, it must do so given access only to noisy, dimensionally-reduced representation odor world provided by receptor neurons. As result, solve compressed sensing problem, relying fact that handful millions possible are present in scene. Inspired this principle, past works have proposed normative models for decoding. However, these not captured unique...
The widespread adoption of deep learning to build models that capture the dynamics neural populations is typically based on "black-box" approaches lack an interpretable link between activity and network parameters. Here, we propose apply algorithm unrolling, a method for learning, design architecture sparse deconvolutional networks obtain direct interpretation weights in relation stimulus-driven single-neuron through generative model. We characterize our method, referred as unrolled (DUNL),...
Summary Time is our scarcest resource. Allocating time optimally presents a universal challenge for all organisms because the future benefits of investments are uncertain. We developed normative framework assessing bounded optimality in allocation, emphasizing accuracy predictions, independent subjective costs and benefits. In common decision task across humans, rats, mice, we varied uncertainty by titrating ambiguous sensory evidence measured each subject was willing to invest...
Reinforcement learning has been successful across several applications in which agents have to learn act environments with sparse feedback. However, despite this empirical success there is still a lack of theoretical understanding how the parameters reinforcement models and features used represent states interact control dynamics learning. In work, we use concepts from statistical physics, study typical case curves for temporal difference value function linear approximators. Our theory...
Abstract For animals to navigate an uncertain world, their brains need estimate uncertainty at the timescales of sensations and actions. Sampling-based algorithms afford a theoretically-grounded framework for probabilistic inference in neural circuits, but it remains unknown how one can implement fast sampling biologically-plausible spiking networks. Here, we propose leverage population geometry, controlled by code dynamics, samplers We first show that two classes samplers—efficient balanced...
Rational decision makers aim to maximize their gains, but humans and other animals often fail do so, exhibiting biases distortions in choice behavior. In a recent study of economic decisions, humans, mice, rats have been reported succumb the sunk cost fallacy, making decisions based on irrecoverable past investments detriment expected future returns. We challenge this interpretation because it is subject statistical form attrition bias, observed behavior can be explained without invoking...