The neocortex is capable of performing universal computations through sequence learning, prediction, and replay. The Hierarchical Temporal Memory (HTM) algorithm is a computational model that embodies these processes. However, the HTM algorithm relies on abstract discrete-time neuron and synapse dynamics, which limits its biological plausibility. In this study, the authors propose a continuous-time implementation of the temporal-memory (TM) component of the HTM algorithm using a network of spiking neurons with biophysically interpretable variables and parameters. The proposed model learns high-order sequences by employing a structural Hebbian synaptic plasticity mechanism supplemented with rate-based homeostatic control. This combination, along with nonlinear dendritic input integration and local inhibitory feedback, enables dynamic self-organization of narrow sequence-specific subnetworks. These subnetworks facilitate the propagation of sparse, synchronous activity, leading to robust and context-specific predictions of future sequence elements as well as autonomous replay of previously learned sequences. By incorporating more biologically realistic features into the TM component of the HTM algorithm, this continuous-time implementation strengthens the link to biology. It allows for better evaluation of the TM hypothesis through electrophysiological and behavioral data analysis by providing a comprehensive exploration of its capabilities in learning sequences and generating predictions. Furthermore, it highlights its potential relevance to understanding neural processing in biological systems. Overall, this study presents an important advancement in bridging computational models with biological mechanisms by developing a more biologically plausible implementation for sequence learning and prediction in networks of spiking neurons.
The neocortex is a part of our brain that helps us think and learn. It does this by remembering patterns and making predictions. The Hierarchical Temporal Memory (HTM) algorithm is a way to mimic how the neocortex works. It uses special math equations to simulate the activity of neurons in our brain. The HTM algorithm can learn and remember sequences of events. Scientists have created a new version of the HTM algorithm that uses real neuron activity to make it more like how our brains work. This new model can help us understand how our brains learn and predict things better."
Definitions
- Neocortex: The part of our brain responsible for thinking and learning.
- Hierarchical Temporal Memory (HTM) algorithm: A mathematical method that imitates how the neocortex works.
- Neuron: A cell in our brain that helps transmit information.
- Synapse: A connection between two neurons where information is passed.
- Dynamics: How things change or move over time.
- Hebbian synaptic plasticity mechanism: A process where connections between neurons become stronger when they are used often.
- Homeostatic control: The ability to keep things balanced or stable.
- Self-organization: When things arrange themselves in an organized way without needing someone to tell them what to do.
- Electrophysiological data analysis: Studying electrical signals in the body, like those produced by neurons, to understand how things work.
- Behavioral data analysis: Studying people's
Exploring the Hierarchical Temporal Memory Algorithm with a Continuous-Time Implementation of Spiking Neurons
In recent years, advances in neuroscience have revealed that the neocortex is capable of performing universal computations through sequence learning, prediction, and replay. To explore this further, researchers have developed computational models such as the Hierarchical Temporal Memory (HTM) algorithm to better understand these processes. However, the HTM algorithm relies on abstract discrete-time neuron and synapse dynamics which limits its biological plausibility. In order to bridge this gap between computational models and biology, a team of researchers from the Institute for Cognitive Science at Université de Montréal recently proposed a continuous-time implementation of the temporal-memory (TM) component of the HTM algorithm using spiking neurons with biophysically interpretable variables and parameters.
A Structural Hebbian Synaptic Plasticity Mechanism
The proposed model learns high-order sequences by employing a structural Hebbian synaptic plasticity mechanism supplemented with rate-based homeostatic control. This combination enables dynamic self-organization of narrow sequence specific subnetworks which facilitate propagation of sparse synchronous activity leading to robust predictions about future elements in learned sequences as well as autonomous replay capabilities.
Nonlinear Dendritic Input Integration & Local Inhibitory Feedback
In addition to structural Hebbian synaptic plasticity mechanisms, nonlinear dendritic input integration and local inhibitory feedback are also employed in order to enable more biologically plausible implementations for sequence learning and prediction in networks of spiking neurons. These features allow for better evaluation through electrophysiological data analysis by providing comprehensive exploration into TM’s capabilities when it comes to learning sequences and generating predictions.
Bridging Computational Models & Biology
Overall, this study presents an important advancement in bridging computational models with biological mechanisms by developing a more biologically plausible implementation for sequence learning and prediction in networks of spiking neurons. By incorporating more realistic features into TM components within HTM algorithms, this continuous time implementation strengthens links between computation models & biology while highlighting potential relevance towards understanding neural processing within biological systems
Sequence learning, prediction and replay have been proposed to constitute the universal computations performed by the neocortex. The Hierarchical Temporal Memory (HTM) algorithm realizes these forms of computation. It learns sequences in an unsupervised and continuous manner using local learning rules, permits a context specific prediction of future sequence elements, and generates mismatch signals in case the predictions are not met. While the HTM algorithm accounts for a number of biological features such as topographic receptive fields, nonlinear dendritic processing, and sparse connectivity, it is based on abstract discrete-time neuron and synapse dynamics, as well as on plasticity mechanisms that can only partly be related to known biological mechanisms. Here, we devise a continuous-time implementation of the temporal-memory (TM) component of the HTM algorithm, which is based on a recurrent network of spiking neurons with biophysically interpretable variables and parameters. The model learns high-order sequences by means of a structural Hebbian synaptic plasticity mechanism supplemented with a rate-based homeostatic control. In combination with nonlinear dendritic input integration and local inhibitory feedback, this type of plasticity leads to the dynamic self-organization of narrow sequence-specific subnetworks. These subnetworks provide the substrate for a faithful propagation of sparse, synchronous activity, and, thereby, for a robust, context specific prediction of future sequence elements as well as for the autonomous replay of previously learned sequences. By strengthening the link to biology, our implementation facilitates the evaluation of the TM hypothesis based on electrophysiological and behavioral data.
