Transfer learning for Time Series Classification

During the last two decades, Time Series Classification (TSC) has been considered as one of the most challenging problems in data mining [2]. This is due to its’ large number of practical applications in various domains such as cyber security, medical, activity recognition, energy and transportation. Stock market anomaly detection in business, identifying heartbeat patterns of patients in hospitals and detecting temperature levels in climate science are some of its’ practical examples. Accurate time series classification can increase the business revenue as well as facilitate optimal resource allocation. Notable algorithms have been developed to address the classification problem, while the vast majority of research has focused on developing similarity measures for accurate classification. Significant challenges face time series classification including the diversity of data that is inherited from the diversity of domains from-where data has been collected. Scarcity of labelled data is one of the most common challenges in TSC. The significant difference between deployment and target domains is another challenge. One way to overcome these challenge is to utilise transfer learning.

Transfer learning is the process of first training on a source domain, and then transferring the learnt knowledge to the target domain [4]. Hence, transfer learning can leverage the already existing data of some related task or domain to understand a new domain. This idea has been shown to improve capabilities of machine learning techniques in many tasks such as computer vision and pattern recognition especially with the advance in deep learning approaches. Transfer learning includes many types which include feature transfer, parameter transfer. meta data transfer and model transfer.

More information on the project, from potential impact to references, can be found on the accompanying PDF.

To apply, please complete the project proposal form and the online application.

Only local Hubs members can access this page. Join the community today: https://archive.phdhub.eu/register