Automated Machine Learning - Auto ML

Automated Machine Learning - Auto ML
Aim

Cooperative learning between man and machine

Benefit

Humans can make the right decision supported by AI

"AI by people for people" is the motto with which EDI GmbH applies artificial intelligence (AI) in the form of automated machine learning, because we always focus on people. Man and machine learn cooperatively from and with each other.

However, machine learning can only be effective and efficient if you formulate the right question in advance. What do you want to get an answer to? What exactly do you want to predict? In a combustion process, for example, this can be the CO2 content, which is then called the core variable. Just as important as the right parameter of the core variable is the determination of the business objective that is to be achieved with the use of AI. We identify both together with the customer in advance in a compact design thinking workshop.

What is special about EDI GmbH's approach to answering the question through AI is the combination of several machine learning algorithms. In the example of recognising elements on technical drawings, this is first an algorithm that recognises the images. Then a second algorithm determines the similarities of the recognised elements. Another crucial point is the correct selection of algorithms that fit the question and the available database, because not every algorithm delivers the same results. The right tool must be used to correctly formulate the question, because not all AIs are the same.

Our strength lies in focusing on the core variable, a single value that is determined by the combination of algorithms and that provides the information that then enables the human or the machine to react. The human makes a decision based on this value, or machines can be controlled by the value so that they are serviced in time, for example, to ensure product quality, increase their lifespan and reduce maintenance costs. This is also called predictive maintenance.  

Top-down and bottom-up, these are the two directions from which we examine the customer's defined questions and which we call the hybrid model.

With the top-down approach, EDI creates models that a human being can understand. This means that the relevant interrelationships are mapped for specific, selected situations. The aim is to be able to distinguish and understand different situations. For this purpose, the knowledge of the experts is formalised and integrated with our patented solution. With this additional knowledge, the existing amount of data is now filtered and clustered by EDI GmbH. Subsequently, local models are generated from this, which are then used to identify and describe individual relevant situations.

The second starting point approaches the question from the opposite direction: bottom-up. The AI is fed with the data based on the previously created models and is then able to recognise patterns and identify situations and data patterns on its own. The network learns itself which patterns are important. It finds answers independently and can also say which pattern in the data was ultimately decisive. 

The insights that the AI delivers with the bottom-up method are then checked again with the top-down approach. This creates a self-improving cycle. The top-down & bottom-up approach of the AI-based hybrid model enables the experts to learn cooperatively with the AI. The crucial aspect of this method is the possibility to better understand the problem. Thus, the view into the future succeeds and a robust answer to the previously defined question is provided.

If you are thinking about using AI in your company, please do not hesitate to contact us. We look forward to hearing from you and discussing your specific question and first steps!