PolyNet – the first tool to automatically train and optimise a neural network; giving you better quality results in A.I. and Machine Learning.

There are many companies and organisations currently using Machine Learning and Artificial Intelligence who have been getting good results, but they have also been getting a little frustrated –  as they feel results could possibly be better.

That’s exactly where we can help.

At the heart of your ML and AI, underlying everything – is the neural network. Until now, finding ways to choose the best configuration and therefore performance of your neural network has been at best, educated guesswork.  The choices are essentially about weights and architecture.

We are introducing a smart tool that gets right to the heart of the problem, automatically exploring and then analysing and computing with  its onboard engine – to improve the structure and performance of the neural network.

Rather than making an educated guess,  which is a manual task, it actually computes the solution. Taking away the elements of hunches and opinions inherent in existing methods.

That is the core difference we utilise to improve accuracy and, at the end, results from your ML and AI.

PolyChord is the maths engine driving the process. It was developed by Astrophysics Professors in Cambridge University in 2017, and has unique abilities to solve high-dimensional data problems. Exactly the problems relating to neural network configuration and behaviour. This same team has specifically adapted it to explore and optimise the neural network with the tool we created by using PolyChord to train the neural network, that tool is called  PolyNet.

PolyNet is a first, a ready-to-use solution which trains your neural network,  which you (meaning your technical team) can deploy in house, so your data never has to leave the building.

Mathematically PolyNet explores the entire space occupied by your neural net and then analyses it.  Other systems only explore part of the space. PolyNet is at the working  stage right now and we are booking demos for the ready-to-use tool in October and November 2018.

There are one or two other neural net training tools in this space already.  TensorFlow users will know that TensorFlow can help you to choose correct weights, for instance.  However it is working on a different principle and that principle can get stuck in one avenue of exploration.  PolyNet however will go through all of the  network exploring the entire space, making smart choices about both weights and structure, then make automated adjustments to optimise performance and functionality – the result for you being improved outcomes. PolyNet may suggest for instance that a more compact approach may produce better results.

That’s the unique and key difference with PolyNet. As far as we know, no other current tool can help you make informed, machine-calculated choices about weights and architecture in this way.

We are currently offering demo’s of the system, as this is the best way for potential users to ask questions about PolyNet, explore the possibilities, and see how improvements can be gained. There has been a lot of interest. This demo is free to qualifying organisations and is given personally to key members of your team by senior personnel from PolyChord Ltd.  A successful demo may lead to trial use, trials are on a paid-for basis over two months.

We have flexible options with which to move forward after a successful trial (including fee-based, or as an alternative for some customers, metrics-driven performance-based with no front end fee). PolyNet is not aimed at small, low-budget users. It is a unique new cutting edge application which can confer a significant competitive advantage to organisations who have already progressed with ML and AI. Contact us to express an interest in a free demo. We are happy to demonstrate to serious enquirers without cost or obligation.

Please tell us some details about the nature of your enquiry

I consent to having this website store my submitted information so that they can respond to my inquiry.
 

See our notes if you running a neural network/AI project for medical charitable purposes.

PolyChord is also the engine inside our ongoing development for 2019, PolyFold, an early stage experimental Protein Folding tool. It uses PolyChord’s unique ability to solve high dimensional problems to give those wanting to unfold proteins a “turnkey” solution.