Recent advancements in sensor technology have encouraged an inexpensive deployment of large-scale wireless sensor networks (WSNs). Recalibration of sensors in large-scale WSNs is a major challenge and affects the quality of monitoring. Various applications requiring blind calibration include environmental monitoring, building structure health monitoring, and precision agriculture. This project aims to propose a self-calibration framework using incremental deep learning for sensor measurements. The project will also consider the development of a graph-based extreme deep learning framework to improve the accuracy of calibration under noisy and uncontrolled environments.
- Model the influence of environmental factors on the calibration/re-calibration capability of sensor networks.
- Propose a suite of incremental/extreme machine learning/deep learning approaches to extract spatial and temporal features to generate drift-free measurements.
- Evaluate the proposed approaches to re-calibrate sensors deployed under uncontrolled environments.
- Develop a graph-based extreme deep learning model to increase the accuracy of calibration under noisy measurements.
- Establish benchmarks on the generalization ability of incremental and graph-based incremental learning under uncontrolled environments.
- Candidate should have Masters in Artificial Intelligence/Big Data Analytics/ Computer Science. Candidate should have at least a 2:1 Honour’s degree, or equivalent, in Computer Science/Mathematics or relevant discipline. Experience in Machine Learning tools (TensorFlow, PyTorch, Keras) and computational modelling is strongly desirable. Ability to pursue independent research and excellent writing and fluency in English are mandatory.
Please send to Sean Walker – [Email Address Removed] only using the application form.
Application Form / Terms of Conditions can be obtained on the website:
The closing date for receipt of applications is 5pm, (GMT) 21st February 2022