ADAGOS drastically reduces the size of artificial neural networks, enabling Continental to embed them in their cars of the future.
ADAGOS has developed a new parsimonious approach that reduces the resources (including energy) required for implementing machine learning algorithms by orders of magnitude. This approach was the missing piece that will enable Continental to complete the puzzle of embedding its artificial intelligence algorithms into the cars of the future; an ambition that requires hunting down and eliminating every possible source of wastage.
The state of the art of artificial intelligence is largely inspired by the biological brain, including its redundant nature. While this redundancy may ensure the continued functionality of the living brain despite regular, and sometimes accidental, loss of the neural cells, the same argument does not hold for artificial neural networks, which are made from inert matter.
Occam’s razor, also known as the principle of parsimony, proposes that from a set of competing hypotheses, the simplest solution is the most credible, or as Einstein phrased it: "Everything should be made as simple as possible, but no simpler." However, making things simple is not always itself a simple task and so the use of redundant neural networks persists in the state of the art.
In accordance with Occam’s razor, ADAGOS configures small, parsimonious neural networks in a fully automatic manner. Meanwhile, the state of the art is still limited by a tedious, manual trial and error process; this manual approach is itself a source of redundancy and cannot produce neural networks of a small size.
Furthermore, many artificial intelligence users, particularly those in the field of the healthcare, complain about their methods yielding bizarre results. Parsimonious neural networks provide an effective solution to this problem; they are able to capture with great accuracy the physical or biological phenomena conveyed by the data. They even allow creation of highly reliable complex dynamical models, including for quasi-chaotic phenomena; a case in which even the slightest flaw of parsimony would have irremediable consequences on the quality of the model.
A major industrial player put ADAGOS in competition with one of the world-leaders in AI on the problem of predicting turbocharger blade deflections. Without a doubt, the results were in our favor.