Demonstrate the use of CMA-ES for training transformers

Longer description

Context And Background: 

SingularityNET Foundation, in collaboration with other partners such as the OpenCog Foundation and TrueAGI, is working toward a scalable implementation of the Hyperon AGI framework running on decentralized infrastructure, and toward implementation of the PRIMUS cognitive architecture within this framework.

Hyperon and PRIMUS are complex systems involving multiple components, which need to demonstrate appropriate functionalities both individually and in combination.  

This RFP aims to address a portion of this overall need, via funding the initial iteration of one significant component of PRIMUS within Hyperon: Exploring and Demonstrating the use of evolutionary methods (EMs) such as Covariance Matrix Adaptation Evolution Strategy (CMA-ES) for training various Deep Neural Networks (DNNs) including transformer networks. 

The primary objectives of this EM exploration and demonstration RFP are to determine the effectiveness of:

• EMs for updating node weights;
• EMs for evolving DNN architectures;
• EMs for creating a potentially new type of “evolutionary DNN”

Ideas/questions:

• Explore the use of CMA-ES for training transformers and other DNNs
• What are its strengths and weaknesses?
• For what DNN problem domains would CMA-ES likely work best?
• CMA-ES strategies for DNNs (e.g. evolution of weights, architectures, both?)
• Integrated LLM/evolutionary methods in new DNN architectures and/or weight updating (LLM for direction/evolution for invention).

Note:

We are seeking innovative and creative ideas of what is meant by an “evolutionary DNN”. In this spirit, we encourage all manner of interpretations about how to integrate evolutionary methods with DNNs.

Collaboration:

This RFP will be followed by subsequent RFPs for applications that leverage Hyperon/PRIMUS to carry out various applications, and that aim to guide Hyperon/PRIMUS systems in cognitive development toward beneficial AGI

RFP Expected Outcomes:

• Initial Deployment of “evolutionary DNNs” in a demonstration Hyperon instance hosted by SingularityNET 
• OSS code: Provision of code underlying the initial deployment in an open code repository with an appropriate OSS license.
• Thorough Documentation: provide comprehensive documentation detailing the structure and functionality of the evolutionary DNNs to allow other teams to work with and to work on the evolutionary DNNs.
• Technical Report: provide a technical report summarizing experiments run with evolutionary DNNs and results obtained. This should be sufficient to allow others to replicate the work, leveraging the open source code made. The report must include a description of the strengths and weaknesses of each evolutionary DNN, comparisons to standard DNNs, and identification of problem domains in which evolutionary DNNs perform best.