Adkodas

About Us


 

Welcome to Adkodas™: a different kind of data analysis system™.

Do you know what your computers are thinking? More to the point, Adkodas can help your research, and aid you to learn your data, but if you're wondering how computers come to a conclusion, you've come to the right place

Adkodas is in the business of a new kind of deep learning, taking computers into machine learning (or AI) -- that is, putting computers to work for you to find useful knowledge in and make predictions from your data -- and beyond. While much of current deep learning uses increased raw computing power or more layered networks to learn from big data, Adkodas uses revolutionary new technology to not only itself learn, but also help you learn what it knows.

Whether you are a newcomer to data analysis/"big data" or an expert looking for an intelligent, systematic machine-driven approach, Adkodas can help you figure out where to start in unlocking the potential of what you've collected.

We live in an era of exponentially growing amounts of data. Raw data, by itself, just occupies space--to derive benefit, academically, financially, or otherwise, you need to understand the patterns hidden within your data. Finding those patterns can be daunting. Any number of services can promise "magic" out of a black box: most AI can train on some data and make oracular, mysterious predictions based on statistical weighting, but all you can know is the results of testing those AI, not necessarily what they are "thinking" or how they arrive at their results.

To demystify the arcane results of deep learning, Adkodas is proud to be able to offer rule extraction: a major difference between our neural networks and others' is the ability to understand what our machines have learned. It's not uncommon for AI to give an unexpected result--and how happy are you to simply trust an emergent technology to get your data right? With Adkodas' rule extraction, you can see why a certain outcome is predicted -- this allows you both to use our AI with confidence and to learn the underlying principles of your data in an independently verifiable way. When you find the rules, or patterns, in your data, you can discover the principles of the research you are doing, and not merely present "some AI told me" as a result. (See an Adkodas research experience!)

This is the power of our advanced rule extraction services.


Adkodas services

One of the key differences between Adkodas tech and others' is that we don't need to add a "twenty questions" translation protocol to try to guess what our networks "might" know - or don't actually know. Like biological learning, our AI isn't statistical; it either knows what it's learned from the data, or it knows that it doesn't know.

More generally, Adkodas can perform a number of data analysis services:

The range of Adkodas' tailorable services is illustrated through freely available analysis of public-domain data sets with examples of our rule-extraction services, as well as more about our rule-extraction and neural-network analysis services (and how they differ from other machine learning).

  • German Credit Problem
    German Credit Problem
  • Abalone Classification
    Abalone Classification
  • Forest Fires Predictions
    Forest Fires Predictions
  • E.Coli Classification
    E.Coli Classification
  • Astronomical Survey Classification
    Astronomical Survey Classification

But... isn't AI dangerous?

In a nutshell, not in itself -- AI is a tool, and so is exactly as dangerous as the uses to which it is put. If the results of AI are simply applied without understanding how it arrived at those results, that is potentially dangerous and takes us from the realm of true learning to consulting a computer oracle with blind faith. If by "dangerous" one means the much-vaunted singularity... well... there is much more to say. In short, that's not the danger; the danger is not knowing what you're actually doing with your data, and that's what Adkodas can help you with.

Coda and what's next

The algorithms we use to process data are based on proprietary intellectual property, public domain, as well as patented and patent pending algorithms; Adkodas$#0153 technology may be covered by one or more of the following patents: United States Pat. No.: US 8,862,527 ; Canada Pat. No.: 2,629,069 ; Japan Pat. No.: 5395241 ; Korea Pat. No.: 10-1326914 ; New Zealand Pat. No.: 567815

Many projects yet lie ahead, including increasing parallel-processing capability for the largest data sets; this generation of Adkodas™ technology is only the beginning!

Legal    Privacy   Contact Us