Ben Halstead, PhD

GitHub LinkedIn Google Scholar

As a machine learning researcher my passion is developing innovative solutions to difficult problems which deliver world-changing value. With experience in both industry and academia, I have the skills to develop and implement cutting edge, production ready solutions by combining deep R&D and software engineering expertise. For my PhD research I developed fundamental algorithms for real-time reactive systems capable of handling changes in the distribution of streaming data. In my professional career I have developed and deployed state-of-the-art approaches to a range of applied problems, including streaming sensor data, satellite powered predictive systems and NLP analysis.

PhD Research

A growing amount of data now comes from streaming data sources, where new data is collected and added in real time, rather than traditional data sets. For example, collecting air quality data in real-time from a network of sensors, or nearly all online businesses which collect information. Traditional machine learning methods are designed around static datasets, making the assumtion that data will stay similar over time. If data changes, for example, shifting weather patterns or consumer behaviour, these static methods can become disastrously out of date.

My PhD research focused on allowing machine learning methods to detect changes, identify the new emerging patterns, and adapt themselves to retain performance. In particular my research enabled systems to identify recurring patterns which can be reused for performance benefits and tracked over time to trace changes in hidden context.