A break from the state-of-the-art#

NeurEco is a unique automatic artificial intelligence software. It is based on an innovative parsimonious artificial neural network (ANN) approach. You simply need to provide data, and NeurEco will automatically build the model that gives the best prediction possible. The produced models are generally significantly smaller than what traditionally thought was possible. There is no tradeoff between model size and accuracy.

NeurEco was designed by engineers to meet the needs of engineers. It is oriented towards technical data. It can easily process both simulation and measured data.

NeurEco considerably reduces computing resources and energy consumption. NeurEco is designed to be used on a single computer and does not require cloud computing capabilities.

NeurEco starts from the smallest possible neural network which is enriched step by step. The intermediate states of the model are available via checkpoint. This process stops when it is not possible to get a better prediction by enriching the neural network.

Our drastic parsimony forces the learning process to infer rules structuring the data for a more intelligent learning and a better prediction.

Unlike in the classical layered neural network structure, a neuron in a NeurEco network can be connected to any other neuron in the network. In our context, a layer is the set of neurons that can be evaluated simultaneously. This structure unlocks the possibility of creating small parsimonious neural networks.

NeurEco provides, alongside the generated neural model, an estimated generalization error that predicts the performance of the neural network on unseen data. The network construction process is deterministic so the created neural networks are reproducible with the same processing unit/OS.