Abstract:We have built a neuronal model of decision making. Our model performs a decision based on an imperfect discrimination between highly mixed stimuli, and expresses it with a saccadic eye movement, like real living beings. We use populations of integrate-and-fire neurons. To take a decision that depends on imperfectly separated stimuli, we use a model inspired by the principle of accumulation of evidence. More precisely, this accumulation of evidence is performed by a competition between groups of neurons that model a visual column in the lateral intra-parietal area (area LIP) of the brain. In this column, we have groups of neurons that are sensitive to specific stimuli on which the decision is based. They inhibit each other through groups of inhibitory neurons. Simultaneously, they recursively excite themselves through recurrent synaptic excitation, and all the neurons receive a steady low-level excitation from the rest of the brain. The competition is generated by these recurrent inhibitory and excitatory loops. We study this structure within the framework of dynamical systems. The variables we use are the activities of each group of neurons. This dynamical system has several stable states: one of them occurs when all activities are weak, and other ones when one local group of neurons has a higher activity and dominates the others through lateral inhibition. The convergence to one of these stable states models decision making, guided by sensory evidence. The group of neurons sensitive to the specific stimulus has a comparative advantage on the others during the competition. This structure is not a new idea, but we use it to test our hypothesis on the way the brain controls the dynamics of decision making. Our hypothesis is that the statistical signature of the low-level activity of the brain modifies the stability of the attractors of our model, and thus changes the dynamics of the competition that models decision making. The criterion by which we judge that a decision is taken is more realistic than just looking at the decisive neurons' activities. We model a saccadic eye movement directed by the activities of our LIP neurons, and we read the decision from the position of the eye. This experimental setup is comparable to biophysical experiments in which living beings express their decisions by saccadic eye movements. The neurons of the LIP column in which the decisions take place are modeled as neurons in a real brain: besides the stimuli and the recurrent interactions, they receive significant inputs from the rest of the brain. It is well known that neurons in the brain are highly influenced by this activity, called low-level background activity. We study how the dynamics of the decision making change as a function of the first-order statistics (mean) and second-order statistics (variance) of this global low-level background activity or noise. By studying its influence on the reaction time and error rate, we show that this background activity may be used to control the dynamics of decision making. We compare the performance of such a model (error rate as a function of reaction time) to the performance of living beings during psychophysical experiments. By doing so, we assess the plausibility of the hypothesis that decisions be controlled by the statistical signature of the low-level background activity of the brain.

Neural Dynamics Model of Visual Decision-Making: Learning from Human Experts

A Neurocomputational Model of Decision and Confidence in Object Recognition Task

Modeling the Human Visual System: Comparative Insights from Response-Optimized and Task-Optimized Vision Models, Language Models, and different Readout Mechanisms

A Neural Network Model of Visual Attention Integrating Biased Competition and Reinforcement Learning

Seeing eye-to-eye? A comparison of object recognition performance in humans and deep convolutional neural networks under image manipulation

A large-scale examination of inductive biases shaping high-level visual representation in brains and machines

Neural Mechanisms of Human Decision-Making

Unidirectional brain-computer interface: Artificial neural network encoding natural images to fMRI response in the visual cortex

RTify: Aligning Deep Neural Networks with Human Behavioral Decisions

Predicting Single Neuron Responses of the Primary Visual Cortex with Deep Learning Model

Neuronal Model of Decision Making

Deep Reinforcement Learning Models Predict Visual Responses in the Brain: A Preliminary Result

A Novel Biologically Inspired Visual Cognition Model: Automatic Extraction of Semantics, Formation of Integrated Concepts, and Reselection Features for Ambiguity

Neural Foundations of Mental Simulation: Future Prediction of Latent Representations on Dynamic Scenes

A Neural Model of the Frontal Eye Fields with Reward-Based Learning.

Brain-inspired Models for Visual Object Recognition: an Overview

Neural Decoding of Visual Information Across Different Neural Recording Modalities and Approaches

Biologically Inspired Visual Model With Preliminary Cognition and Active Attention Adjustment

Biologically inspired model simulating visual pathways and cerebellum function in human - Achieving visuomotor coordination and high precision movement with learning ability

Achieving More Human Brain-Like Vision via Human EEG Representational Alignment

Modelling Human Visual Motion Processing with Trainable Motion Energy Sensing and a Self-attention Network