Search

Riveland and Pouget model instructed action, showing that shared structure in task and semantic representations allows language to compose practiced skills in novel settings. Models make predictions for neural activity in human language areas.

The International Brain Laboratory presents a brain-wide electrophysiological map obtained from pooling data from 12 laboratories that performed the same standardized perceptual decision-making task in mice.

Individual dopaminergic neurons encode future rewards over distinct temporal horizons.

Focused grass-roots collaborations that start small and scale up could overcome technical and sociological barriers to 'big' neuroscience, argue Zachary F. Mainen, Michael Häusser and Alexandre Pouget.

Brain-wide recordings in mice reveal that prior expectations are distributed through recurrent loops across all levels of cortical and subcortical processing.

Neuroimaging modalities such as MRI and EEG are able to record brain activity, but spatiotemporal resolution and sensitivity are limited. Here, the authors show how a recently developed method, functional ultrasound imaging (fUS), can measure brain activation during cognitive tasks in primates.

Here, the authors show that rats’ performance on olfactory decision tasks is best explained by a Bayesian model that combines reinforcement-based learning with accumulation of uncertain sensory evidence. The results suggest that learning is a critical factor contributing to speed-accuracy tradeoffs.

We propose that synapses compute probability distributions over weights, not just point estimates. Using probabilistic inference, we derive a new set of synaptic learning rules and show that they speed up learning in neural networks.

Everyday decisions require choosing among multiple options. This work derives the optimal decision policy and shows how it can be approximated by a biologically plausible neural circuit and how this circuit can reproduce observed behavior.

Drift diffusion models (DDM) are fundamental to our understanding of perceptual decision-making. Here, the authors show that DDM can implement optimal choice strategies in value-based decisions but require sufficient knowledge of reward contingencies and collapsing decision boundaries with time.

The authors show that a normative approach to olfaction, Bayesian inference, reproduces much of the anatomy, physiology and behavior seen in real organisms. The model provides insight into how the olfactory system demixes odors, and, by extension, how other sensory systems extract relevant information from activity in peripheral organs.

Correlations of noise in neural population activity are thought to limit the amount of information contained in such population activity, whereas decorrelation is suggested to increase information content. Here the authors show that decorrelation does not imply an increase in information, and only certain types of correlations limit information content.

Combined intraocular injection of an adeno-associated viral vector, encoding an optogenetic sensor, with light stimulation via engineered goggles enables partial recovery of visual function in a blind patient.

Noel et al. show aberrant updating of expectations in three distinct mouse models of autism spectrum disorder. Brain-wide neurophysiology data suggest this stems from excess units encoding deviations from prior mean and a lack of sensory prediction errors in frontal areas.

The authors implement model-based analyses to uncover strategies used by mice and humans during sensory decision-making. Contrary to common wisdom, mice do not lapse and, instead, switch between sustained engaged and disengaged states.

Lightning Pose is an efficient pose estimation approach that requires few labeled training data owing to its semi-supervised learning strategy and ensembling.

Perceptual learning has been proposed to result from improvements either in early sensory processing or at the later stage of sensory decoding. Here the authors show that altering the feedforward connectivity in a recurrent neural network so as to improve probabilistic inference in early visual areas results in both modest changes in tuning curves and reduced noise correlations.

Using a multisensory cue-conflict task, the authors report that monkeys employ the optimal strategy of weighting each cue in proportion to its reliability, and that population decoding of neural responses from area MSTd predicts behavioral cue weighting. This behavior is further linked to the specific computations by which single neurons combine their inputs, consistent with recent theories of optimal probabilistic neural computation.

When making a decision, we have to take into account not only the information that is available, but also the reliability of this information. This behavioral study finds that, while searching for a visual target, people weigh up cue reliability in an almost identical fashion to a mathematical ideal observer, and a neural network model can explain how this behavior is produced.

Correlations in firing rate between pairs of neurons can change depending on task and attentional demands. This new finding suggests that measuring correlations can help to reveal how neural circuits process information.

What is the minimal sensory processing time before we can make a decision about a stimulus? A study now reports that, for simple perceptual decisions, this can take as little as 30 ms.

Understanding how realistic networks integrate input signals over many seconds has eluded neuroscientists for decades. Koulakov and colleagues now propose a computational model to explain how bistable neurons might allow a network to integrate incoming signals.

Computational neuroscientists have started to shed light on how probabilistic representations and computations might be implemented in neural circuits, and here the authors review the application of these theories thus far. They further discuss the challenges that will emerge as researchers start expanding their use to more sophisticated, real-life computations.

The authors use recent probabilistic theories of neural computation to argue that confidence and certainty are not identical concepts. They propose precise mathematical definitions for both of these concepts and discuss putative neural representations.

Sensory and motor information in the brain is represented as activity in populations of neurons. But how does correlated noise affect population coding? These authors evaluate empirical and theoretical evidence on the interactions between correlations, population codes and neural computations.

A modular architecture for managing and sharing electrophysiology, behavior, colony management and other data has been built to support individual laboratories or large consortia.

Gas chromatography is a useful tool to identify and characterize wines, usually by selecting some compounds for a particular classification problem, yet, with limited success. Here, the authors decode the estates perfectly and age 50% correctly of twelve red Bordeaux wines from unrestricted, raw gas chromatograms using machine learning.

One of the ambitions of computational neuroscience is that we will continue to make improvements in the field of artificial intelligence that will be informed by advances in our understanding of how the brains of various species evolved to process information. To that end, here the authors propose an expanded version of the Turing test that involves embodied sensorimotor interactions with the world as a new framework for accelerating progress in artificial intelligence.