Petrofinde’s Extreme Learning Machine (ELM) is a single-hidden layer feed-forward neural network (SLFN). Training SLFN consists of randomly generating the hidden layer weights, followed by solving a linear system of equations by least-squares for the estimation of the output layer weights. The learning strategy is very fast and gives a significantly higher prediction accuracy. Theoretically and practically, this algorithm can produce excellent generalisation performance and can learn thousands of times faster than conventional popular AI learning algorithms for feed-forward neural networks.
Petrofinde is developing a kernel-based ELM (KELM), where the hidden layer feature mapping is determined by the kernel matrix. In this version, only the kernel function and its parameters are needed; the number of hidden nodes is not required. With the use of kernel function, KELM is expected to achieve better generalisation performance than basic ELM. Furthermore, as randomness does not occur in KELM, the chance of result variations could be reduced.
Petrofinde in collaboration with ESA (European Space Agency) and Nvidia, will be dealing with the live treatment of satellite images. This can stack up to several tera-bytes a day, to transfer, process and analyse. The Copernicus project and Sparkin data, analyse land movement from satellite images, which could have an impact on mining, civil engineering and in the oil and gas industry. Thanks to ELM deep learning, It will evolve quickly in the domain of geoscientist assistance. The same goes for production analysis, and text mining.