Announcing: The Emergent Neural Network Simulation System http://grey.colorado.edu/emergent/index.php/Main_Page
Emergent is a major rewrite of the widely used PDP++ system.
Emergent is a comprehensive simulation environment for creating complex, sophisticated models of the brain and cognitive processes using neural network models. These networks can also be used for all kinds of other more pragmatic tasks, like predicting the stock market or analyzing data. Emergent includes a full GUI environment for constructing networks and the input/output patterns for the networks to process, and many different analysis tools for understanding what the networks are doing. It has a new tabbed-browser style interface (in Qt 4), with full 3D graphics (via Open Inventor/Coin3D), and powerful new GUI programming tools and data processing and analysis capabilities. It supports the same algorithms as PDP++: Backpropagation (feedforward and recurrent), Self-Organizing (e.g., Hebbian, Kohonen, Competitive Learning), Constraint Satisfaction (e.g., Boltzmann, Hopfield), and the Leabra algorithm that integrates elements of all of the above in one coherent, biologically-plausible framework.
Relative to PDP++, the main advances are:
- Much easier to modify and extend the "scripting" of network training through a new GUI-based programming system -- everything is transparent and user-modifiable. Considerable support is included for implementing complex psychological tasks via this programming environment.
- The tabbed browser allows everything to be contained within a single window, with full search functions, cut/copy/paste, drag-and-drop, etc, for a modern, highly efficient working environment.
- Everything has been boiled down to the most basic, general-purpose elements, which can now be combined in more powerful, "emergent" ways. Environments and monitor data and all other forms of data have been consolidated in a single powerful DataTable object that supports many different kinds of operations (e.g., database-style Joins and Sorts, vector and matrix math, 3d graphing, statistics, etc). With convenient interfaces for DataTables in the GUI programming environment, flexible and efficient data processing and analysis functions can be readily performed.
- Has a greater variety of network visualization tools, and a built-in virtual environment simulator (based on the popular ODE toolkit) allows networks to interact with a realistic simulated environment, to explore more embodied and robotic functionality.
- Standard GPL license, ./configure build process, native look-and-feel on all 3 major platforms (Linux, Mac, Windows), easily-installable binary packages (including apt & yum on linux), and dynamically-loadable plugin modules.
Relative to the prevalent use of MATLAB and other general-purpose tools for neural neural network simulation, emergent offers several important advantages:
- completely open source, free software.
- highly optimized execution speed, including distributed memory computation, while also supporting complex biologically-based neural architectures.
- designed specifically to make research simulations easily accessible to other users with minimal additional effort: built-in documentation system, pervasive comment fields, accessible, transparent interface.
In brief, if you're doing large scale, complex neural network models, emergent offers many advantages.