Search papers, labs, and topics across Lattice.
This study introduces a novel method for real-time monitoring of high-dimensional time-series data by integrating topological data analysis (TDA) with machine learning techniques. By representing multivariate time-series as manifolds and employing topological descriptors, the authors effectively capture the dynamic evolution of the system using a neural ordinary differential equation. The results demonstrate that this trajectory-based event detection outperforms traditional reconstruction-based methods like PCA and autoencoders, as well as existing trajectory-based approaches such as Koopman autoencoders, in identifying various event types in industrial processes.
A trajectory-based event detection method using topological data analysis outperforms traditional techniques in monitoring high-dimensional dynamic processes.
Real-time process monitoring requires methods that extract actionable information from high-dimensional time-series data. In this work, we present a new approach for process monitoring that combines tools of topological data analysis (TDA) and machine learning. In the proposed approach, we represent multivariate time-series data as manifolds and use topological descriptors to summarize the structure of such data; we then use a neural ordinary differential equation to learn the dynamic evolution of the topological structure of the system. Using real data from an industrial process, we show that this trajectory-based event detection approach is effective at detecting diverse types of events. We contrast this approach against reconstruction-based approaches such as principal component analysis and autoencoders and against a trajectory-based approach that uses Koopman autoencoders.
