Machine learning algorithms build a mathematical model that gives systems the ability to automatically learn and improve from experience without being explicitly programmed. Systems designed with machine learning algorithms can operate without human intervention or assistance and adjust actions accordingly.

In the nuclear industry, for example, Aposys devices use machine learning to identity structural anomalies in the toolsets. The determination of differences in geometry is fundamentally a machine learning process:  the system must determine what threshold is needed to correctly characterize the differences in the structure;. it must successfully reject inconsequential differences such as surface blemishes and measuring noise, and correctly detect missing pieces. 

The superiority of our product stems from the creation of convolutional neural networks (CNNs), which can take an image as an input and then output a desirable parameter. In this case, the output is whether a deviation is detected. Unlike traditional geometry matching, CNNs can directly use the photos that are produced for photogrammetry, and thus serve as an independent check of the existence of missing features. We have been using a shallow network, where we can directly implement a feature map to quickly identify small regions of interest, and perform detection on those.