News
2024-10-03
I have successfully defended my thesis and have been awarded my PhD in Interactive AI!
2024-09-19
Published Euclid preparation - XLIII. Measuring detailed galaxy morphologies for Euclid with machine learning in Astronomy & Astrophysics!
2024-06-03
I have started my 2-year Fellowship at the University of Bristol working on Diffusion Models and Digital Twins!
2024-02-15
Published Time-Series Classification for Dynamic Strategies in Multi-Step Forecasting on Arxiv!
2024-02-15
Presented a talk on 'Optimizing Data Efficiency: Using Active Learning Strategies and the QUEST Method for Efficient Classification and Labeling in Large Datasets' at the Galaxies & AGN with the First Euclid Data and Beyond in Bologna.
2023-03-28
Presented a pecha-kucha talk on my PhD research at the Interactive AI CDT Spring Research Conference.
2022-12-06
I recently finished my 6 month placement as an AI Research Engineer at Imagination Technologies where I was working on Sparsity, Lidar data and CUDA implementations of custom DL layers. A patent for my work has been submitted and is awaiting approval.
2022-04-27
Presented a talk on 'Using active learning to create reliable and robust classifiers for Euclid' at the 2022 Annual Euclid Consortium Meeting in Oslo.
2022-03-30
Presented a pecha-kucha talk on my PhD research at the Interactive AI CDT Research Showcase.
2022-02-23
From June, I will be starting a 6 month placement as an AI Research Engineer at Imagination Technologies!
2021-12-16
Presented a talk on 'Using active learning to create reliable and robust classifiers for Euclid' at the 2021 Euclid Consortium UK Meeting.
2021-10-18
Presented research poster of AstronomicAL at the 2021 IAP Colloquium which was dedicated to critical analysis of Machine Learning methods in Astronomy.
2021-09-03
Published AstronomicAL: an interactive dashboard for visualisation, integration and classification of data with Active Learning in Journal for Open Source Software!
Keyword: multistep-forecasting
Time-Series Classification for Dynamic Strategies in Multi-Step Forecasting
R. Green, G. Stevens, T. de Menezes e Silva Filho, Z. Abdallah
Arxiv Preprint
PDF DOI BIB ABSTRACT Keywords: time-series, multistep-forecasting, regressionMulti-step forecasting (MSF) in time-series, the ability to make predictions multiple time steps into the future, is fundamental to almost all temporal domains. To make such forecasts, one must assume the recursive complexity of the temporal dynamics. Such assumptions are referred to as the forecasting strategy used to train a predictive model. Previous work shows that it is not clear which forecasting strategy is optimal a priori to evaluating on unseen data. Furthermore, current approaches to MSF use a single (fixed) forecasting strategy. In this paper, we characterise the instance-level variance of optimal forecasting strategies and propose Dynamic Strategies (DyStrat) for MSF. We experiment using 10 datasets from different scales, domains, and lengths of multi-step horizons. When using a random-forest-based classifier, DyStrat outperforms the best fixed strategy, which is not knowable a priori, 94% of the time, with an average reduction in mean-squared error of 11%. Our approach typically triples the top-1 accuracy compared to current approaches. Notably, we show DyStrat generalises well for any MSF task.@misc{green2024timeseries, title={Time-Series Classification for Dynamic Strategies in Multi-Step Forecasting}, author={Riku Green and Grant Stevens and Telmo de Menezes e Silva Filho and Zahraa Abdallah}, year={2024}, eprint={2402.08373} }