DENSER selected for Finalist in Humies 2018

The DENSER project has been nominated as finalist in the Human Competitive Awards “Humies” 2018, to be held in Tokyo, at the Genetic and Evolutionary Computation Conference (GECCO).


The Humies awards aim to reward human-competitive results that are the outcome of any form of genetic and evolutionary computation. DENSER seeks to automatically generate Deep Artificial Neural Networks, without the need for human intervention.


The results on the conducted experiments show that DENSER is able to discover Convolutional Neural Networks (CNNs) that are human-competitive on the classification of the CIFAR-10. More surprisingly, the models discovered for the CIFAR-10 are robust, generalise, and scale; i.e., without further evolution they are effective on the classification of the MNIST, SVHN, and CIFAR-100. To the best of our knowledge, the reported CIFAR-100 results are among the best for human and automatically-designed CNNs.


More information about the competition and the remaining finalists can be found in