An Improved Sparse Distributed Memory for Anomaly Detection Applications

点击翻译按钮获取中文摘要

In brief, our industrial partner has a proven Sparse Distributed Memory (or SDM) based technology capable of Anomaly Detection of streaming data, but it requires great amounts of memory, operates slowly and requires manual optimization at setup. What we propose is to design, implement and test feasible solutions, which would alleviate these problems and hence, widen the real-world applicability of their technology.These enhancements will boost MindMick's business value, as it will allow it to apply anomaly detection to GPS tracking data for security & safety applications (e.g., real-time monitoring of Alzheimer patient behaviour). It also allows MindMick to envision embedded software devices, to be used in industrial applications (e.g., pre-emptive maintenance alerts for critical machine parts). Neither of these directions are possible without the computational speed-up and dynamic memory allocation enhancements that we plan.