corticalcomputer / dxnn2 Goto Github PK
View Code? Open in Web Editor NEWTopology and Parameter Evolving Universal Learning Network platform DXNN MK2
Home Page: www.DXNNResearch.com
License: Apache License 2.0
Topology and Parameter Evolving Universal Learning Network platform DXNN MK2
Home Page: www.DXNNResearch.com
License: Apache License 2.0
DXNN MK2 Erlang, By Gene Sher http://DXNN.org http://DXNNResearch.com [email protected] Brief documentation for this package is included in this README file. ------------- 1. LICENSE ------------- Copyright (C) 2009 by Gene Sher, DXNN Research Group, [email protected] All rights reserved. This code is licensed under the Apache License, Version 2.0. Please see the LICENSE file that accompanies this project for the terms of use. --------------------- 2. USAGE and SUPPORT --------------------- I hope that this software will be a useful starting point for your own explorations in the creation of Computational Intelligence. The software is provided as is; however, I will do my best to maintain it and accommodate suggestions. If you want to be notified of future releases of the software or have questions, comments, bug reports or suggestions, send an email to [email protected], or request a user account at www.DXNNResearch.com Alternatively, you may post your questions on dxnn.org or dxnnresearch.com The following explains how to use DXNN. INTRO ----- DXNN is a fully distributed Topology and Weight Evolving Artificial Neural Network system created and invented by Gene Sher. Originaly introduced in the publication preprint of 2010 available on arXiv:1011.6022v3 The current state of the project is composed of the following: Cortex synchronizes Neurons, Sensors, and Actuators. A population monitor controls a population of agents, it is the population_monitor process that spawns agents, waits for them to be evaluated, and then performs the selection, replication, and variation (application of various mutation operators). Furthermore, due to this being a memetic algorithm based system (although can be changed to genetic, by simply switching max_attempts variable to 1), each NN based agent system comes with an exoself process, which performs the synaptic weight tuning. Scapes are simulations, not necessarily of 3d environments. It is through scapes that problems are presented to the agents, and it is the agent's morphology which defines the agent's sensors and actuators, and thus defines which scapes it can interface with. The system itself, the mnesia database, and the top most system, is called polis (It's Greek for city state). This version is 1.0, and it does not have a lot of comments (or any), since I built it primarily for myself. But v2.0, which I will release very soon after this, is better and cleaner implemented. Has more functionality, and is fully documented. It is also the version built inside my upcoming book: Neuroevolution Through Erlang, to be released towards the end of this year (just recently submitted my manuscript to my editor). But if you want to give this version a try (feel free to send me an email if you can't get it to work), then by all means go for it. You might find strange comments somewhere in the source code, since I did not take out any notes I wrote for myself. STARTING POINT -------------- From inside Erlang. 1. %%%%Compilation%%%% make:all(). 2. %%%%Initialize All Databases%%%% First create a folder called “benchmarks”, the system expects it to exist, and writes files to it when performing benchmarks. polis:create(). % This creates the database 3. %%%%Start The Polis Databases%%%% polis:start(). % This starts the polis process, the whole thing, the infrastructure (it runs the scapes...) 4. At this point you can summon NN based agents or populations of agents, construct Sensors and Actuators and provide them to the NNs... This section will be expanded in future additions. To set the population and agents to the preferred sensors and actuators, modify the INIT_CONSTRAINTS (in population_monitor module), and use the particular morphology you want (check the morphology module, different morphologies are for different problems), and then execute population_monitor:start(), which will create the population of size decided by you of agents using the specified morphology and thus the sensors and actuators. If you want the agents to discover and explore the available sensors and actuators (perform feature selection in a sense) then modify modular_constructor, ensuring that you use the get_InitSensors() and get_InitActuators() function, instead of the get_Sensors() and get_Actuators() function used within the module. The get_Init.. starts the population off with the NN based agents using just a single sensor and actuator, exploring other available sensors and actuators within the morphology as they evolve. You can add new sensors and actuators by specifying those sensors and actuators in the morphology module, and then creating those functions in the sensors and actuators modules. New mutation operators, activation functions... all can be added within the records.hrl, as long as those functions are realized/implemented in their respective modules so that they can be executed when called upon.
Not sure if you provided the latest version of fx.erl? Error when initiating forex_trader... can you provide the latest fx.erl file?
Great work, I'm reading your book and testing around.
I'm just curious, from experience, when such systems have bugs, they don't show it, All numbers.
What is the logic behind the decision of not testing?
I don't usually write tests, can't defend it, bad habit, but I wonder why don't you.
derivatives.erl
indicates that it is GPL'd. Did you intend to license it separately?
https://github.com/CorticalComputer/DXNN2/blob/master/derivatives.erl#L7
Hello,
circuit.erl fails on compilation ( on erlang 17 ) ( after a long series of warnings ). Any idea ?
Not an issue per se so feel free to close it out but thought I would share this project https://github.com/Rober-t/apxr_run which is an updated and refactored version of this repo. The credit for the good stuff still goes to the original author but perhaps others will find the project above easier to work with :)
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.