This patent introduces a novel biologically inspired universal generative modelling approach called a Hierarchical Spatial-Temporal State Machine. The patented approach is developed on the understanding of the brain, its structure and functionality. The technique is capable of modelling and predicting complex spatial-temporal patterns in data from which it is able to predict the future states of a system based on its previous behavior while taking into account significant noise in the data. The approach can automatically learn complex real world patterns to identify abnormal conditions. This gives it a competitive advantage over rival methods where substantive human supervision is required.
Due to its unique capability for handling data invariances, our method is able to handle a broad range of data types to discover patterns which are too complex to identify by humans or standard machine learning techniques. The model can be applied to a number of application areas requiring modelling and prediction of complex multi-dimensional data sources in order to find correlations between different inputs (spatial) and how those different inputs vary in time (temporal). The patented model was implemented in computer software and integrated into a number of manufacturing process automation solutions. It has been successfully applied to a number of automotive industrial systems requiring robust and efficient prediction and diagnosis.
The model has a proven record of successful performance in complex data centric manufacturing processes. The performance of the model was compared with a number of rival methods (e.g., deep learning approaches, template based methods, rule and Bayesian based techniques) and proved its superior generative and predictive performance.