A Clojure library implementing Hierarchical Temporal Memory (HTM) for temporal sequence learning and prediction.
Comportex is a Clojure library that implements Hierarchical Temporal Memory (HTM), a neuroscience-inspired theory for temporal sequence learning, prediction, and anomaly detection. It provides a flexible, library-oriented approach to building and experimenting with HTM simulations, allowing users to control simulations and interpret outputs like active cells for custom applications.
Clojure developers and researchers interested in neuroscience-inspired machine learning, temporal sequence modeling, and experimenting with HTM theory for applications like prediction and anomaly detection.
Developers choose Comportex for its pure Clojure implementation of HTM, offering a library-focused design that prioritizes flexibility and user control over automated frameworks, and for its role in understanding and evolving HTM theory through hands-on experimentation.
Hierarchical Temporal Memory in Clojure
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Designed as a library rather than a framework, it gives users full control over simulations and output interpretation, as emphasized in the README, allowing for custom experimentation without rigid constraints.
Aims to help understand HTM theory through visual tools and demos, with resources like essays and the Sanity notebook for interactive exploration, making it ideal for learning and research.
Not a port of NuPIC; built separately based on the CLA white paper and evolved, offering a fresh, Clojure-centric approach to HTM that may inspire new insights.
Leverages Clojure's interactive development with REPL workflows and Leiningen, as shown in the sample workflow documentation, enabling rapid prototyping and experimentation.
The README explicitly states it is 'not yet stable,' meaning breaking changes are likely, and it lacks the reliability needed for production deployments or long-term projects.
Users must manually interpret outputs like active cells to generate predictions or anomaly scores, adding complexity and requiring deeper HTM knowledge compared to frameworks that automate these tasks.
Documentation is limited to essays, wikis, and demos, which may not provide comprehensive guidance for beginners or cover advanced use cases, relying heavily on user experimentation.