Home    Application    FNN    WBN    Contact us   

A new breed of
artificial intelligence

a new breed of
artificial intelligence

Introducing the Artificial Intelligence FNN, a truly unique and entirely new way of taming big data.

FNN – Fleek Neural Network (patent pending) lets you tackle virtually unlimited amounts of data – with less hardware and compute power than ever before! FNN takes a radically different approach to traditional neural networks that does not mute irrelevant inputs via classical backpropagation (~O(n5)) Instead, FNN takes ALL inputs into consideration by using our own method (<O(n2)). , while slashing the hardware requirements to a fraction of other Artificial Neural Networks.

Unlimited data. Endless possibilities.

Text, in volume

Every word and every phrase is an input. Millions of them. We train FNN on huge bodies of text to detect fake news, classify and understand the meaning of texts, and predict the helpfulness of reviews. Do you need to filter entries in your forum or blog that don’t comply with your posting policies? FNN does not simply delete entries containing keywords – but understands the meaning of texts, and filters only those defined as “undesirable content,” doing the job automatically and intelligently and without any manual effort. FNN supports whatever classification you need.

Weather

You want to predict the weather for the next 100 days, taking thousands and thousands of meteorological stations across the globe into account – with amazing accuracy? No problem! We can train FNN to consider millions of inputs about the air pressure in New Zealand and combine the data with the number of sunny days in Alaska. Or any other data, of course. After training to understand what you’re looking for, FNN will deliver unrivaled results without demanding an unaffordable hardware setup

Finance

Maybe you want to evaluate millions of different factors in your prediction of financial behavior? Do you need the headlines of newspapers in India correlated with the average humidity in New York City, and the exchange rate of Swedish Krona to the Malaysian Ringgit to be included? No problem at all; FNN enables going big on data and can learn to provide whatever you need to know. You will be excited about the possibilities – and amazed by the results.

Warehouse

Improve your resource and warehouse management. Predict required stock levels based on the data of your historic sales, in combination with customer wins your company had in the past 47 days (or months, or years…), plus all external data available. Do you want to consider the stock of all the thousands of other products on each day of the last months in your prediction? Let us help you train FNN to deliver exactly what you need.

Medical

Do you want to predict the function and behavior of hundreds of thousands of proteins, each consisting of thousands of atoms? We do it. Or analyze huge epidemiological datasets? You can train FNN on complex or weak structured biological and medical data.

Custom data

Every company is collecting data to optimize processes. In marketing as well as in customer relations and so on... But what if you could get different and better results to optimize your processes, by using your data with FNN. If you want to know if there is a correlation between the time spent on a website and the flow of behavior or the age of the user - we know the answer.

Text, in volume

Every word and every phrase is an input. Millions of them. We train FNN on huge bodies of text to detect fake news, classify and understand the meaning of texts, and predict the helpfulness of reviews. Do you need to filter entries in your forum or blog that don’t comply with your posting policies? FNN does not simply delete entries containing keywords – but understands the meaning of texts, and filters only those defined as “undesirable content,” doing the job automatically and intelligently and without any manual effort. FNN supports whatever classification you need.

Weather

You want to predict the weather for the next 100 days, taking thousands and thousands of meteorological stations across the globe into account – with amazing accuracy? No problem! We can train FNN to consider millions of inputs about the air pressure in New Zealand and combine the data with the number of sunny days in Alaska. Or any other data, of course. After training to understand what you’re looking for, FNN will deliver unrivaled results without demanding an unaffordable hardware setup

Finance

Maybe you want to evaluate millions of different factors in your prediction of financial behavior? Do you need the headlines of newspapers in India correlated with the average humidity in New York City, and the exchange rate of Swedish Krona to the Malaysian Ringgit to be included? No problem at all; FNN enables going big on data and can learn to provide whatever you need to know. You will be excited about the possibilities – and amazed by the results.

Warehouse

Improve your resource and warehouse management. Predict required stock levels based on the data of your historic sales, in combination with customer wins your company had in the past 47 days (or months, or years…), plus all external data available. Do you want to consider the stock of all the thousands of other products on each day of the last months in your prediction? Let us help you train FNN to deliver exactly what you need.

MEDICAL

Do you want to predict the function and behavior of hundreds of thousands of proteins, each consisting of thousands of atoms? We do it. Or analyze huge epidemiological datasets? You can train FNN on complex or weak structured biological and medical data.

Custom data:

Every company is collecting data to optimize processes. In marketing as well as in customer relations and so on... But what if you could get different and better results to optimize your processes, by using your data with FNN. If you want to know if there is a correlation between the time spent on a website and the flow of behavior or the age of the user - we know the answer.

Our Artificial Intelligence FNN (Fleek Neural Network)

Yes, it really works – and we can prove it. Our algorithms are slashing hardware requirements by up to 85%! If you want 20 million different input signals to be considered in deep learning – that’s no problem. We merge big data and neural networks. Requiring significantly less hardware and compute power, FNN makes even the most complex  Artificial Intelligence deployments possible and economical.

Requiring significantly less hardware and compute power, FNN makes even the most complex AI deployments accessible to anyone.

WBN (Well Behaving Networks)

Our WBN (Well Behaving Networks) technology (patent-pending) are deep reinforcement learning networks – an additional “layer of intelligence” to FNN’s, if you will. Providing unique skills such as automatic problem solving and intelligent planning of tactics and scenarios, WBN’s take the past and future into account, tackle problems like object permanence and temporarily hidden objects and much more. We allow artificial intelligence to make the right decisions for the best long-term outcome. 

"IImagine being able to have a look to the future only by analyzing current data - wouldn´t that be amazing?"

The FNN technology

Instead of the classic (~O(n5)), FNN uses our own method (<O(n2)), opening entirely new possibilities for various applications, and making even most complex Artificial Intelligence applications accessible for companies who couldn’t afford it due to the high hardware requirements!

Unlike other ANN’s, our FNN technology does not waste any computational resources to mute “irrelevant inputs” but uses all the data available.  Therefore, FNNs can be “fed” many more input signals – not just a few thousand but millions. This allows the usage of neural networks on really big data.

Some background on Backpropagation

We know that the neural networks in the human brain function like ultimate “super computers”, while common artificial neural networks (ANN) are comparably inefficient. As every additional input signal requires exponentially more resources, most of the computational resources are used to mute inputs classified as “irrelevant,” using classical backpropagation (~O(n5)). For example, increasing the amount of input signals fed to a neural network from 1000 to 2000 input signals, doesn’t require twice the computing power but 32 times as much to handle the request (~O(n5)). Hence, increasing the amount of inputs not just by factor 2 but by factor 100 would require 10 billion times more hardware resources – while still using only a small amount of the available data to do the job.

FNN is radically different, as we have succeeded in developing a new algorithm for backpropagation (~O(n2))! This groundbreaking new approach not only slashes the hardware requirements significantly, but takes exponentially more data inputs into account, making the results more accurate and meaningful than any other Artificial Intelligence solution available today.

The WBN technology

Well Behaving Networks (WBNs) are a technology we developed to provide human-like problem-solving abilities to machine learning. 

This deep reinforcement learning technology consists of four interconnected neural networks and a meta neural network. Together they form a kind of feedback loop inside and outside with the world they interact, which not only enables them to adapt their interactions to react to stimuli but also to simulate the outside part of their feedback loop in their inside loop. 

As a result, WBNs can play and plan scenarios in their networks to get the desired outcome, for example doing something undesirable (like swimming through a cold river) to achieve something desirable (the fruit on the tree on the other side of the river). Or to avoid doing something desirable to avoid something undesirable (like a clever mouse, not eating the cheese in a trap). 

WBNs also develop an understanding of object permanence and temporarily hidden objects and can deduce the past from current information, and in a second step use historical information to enrich current information. 

About us

Fleek Technology Corp.
c/o Aprio, LLP
2002 Summit Boulevard
Suite 120 Atlanta,
GA 30319-1498
F-Enterprises GmbH
Wendelsteinstr. 23
85646 Anzing
Germany

About us

Fleek Technology Corp.
c/o Aprio
2002 Summit Blvd NE Suite 120,
Atlanta, GA 30319
F-Enterprises GmbH
Wendelsteinstr. 23
85646 Anzing
Germany