News

Patent: issued!
Our USPTO patent application for adding a layer of explainability to neural networks without compromising their accuracy has now been issued as well!
READ MORE
Patent: granted!
Our USPTO patent application for adding a layer of explainability to neural networks without compromising their accuracy has been approved!
READ MORE
This Croatian startup is challenging Google’s benchmarks in Deep Learning
“One of Google’s main competitors when it comes to its deep learning models comes from the SEE region. The Croatian startup airt develops a deep learning platform to help companies predict their end customers’ behavior.”
READ MORE
Constrained Monotonic Neural Networks
Our experiments show that our approach of building monotonic deep neural networks have matching or better accuracy when compared to other state-of-the-art methods such as deep lattice networks or monotonic networks obtained by heuristic regularization. This method is the simplest one in the sense of having the least number of parameters, not requiring any modifications to the learning procedure or post-learning steps.
READ MORE
Another patent application: adding a layer of explainability to neural networks without compromising accuracy!
We managed to find a simple solution to a problem that exists in the literature for over 30 years, and which ensures the monotonicity of deep learning models. Compared to the Google’s Deep Lattice Networks that are current benchmark, our solution is more accurate while using at least 10 times less parameters, thus dramatically reducing computing resources necessary for training and predictions.
READ MORE
Top fintech Croatian companies in 2022
We may no longer be “just” a fintech, but it’s still nice to read the following: “The airt team, lead by Davor Runje and Hajdi Cenan, works on advanced technologies and new approaches to analyzing and explaining large datasets. Armed with patents and innovations - the insiders say there’s a bright future in front of airt.”
READ MORE