Horizon 2020 Prize Announced: Call for Papers for follow-up Volume on Spatiotemporal Forecasting

Authors are invited to submit original material to The Springer Series on Challenges in Machine Learning on the subject of “Machine Learning for Spatiotemporal Forecasting” by 17th March 2019. This Volume follows the recent Big Data Technologies Horizon Prize competition organized by the European Commission: https://ec.europa.eu/research/horizonprize/index.cfm?prize=bigdata

Volume Editors: Florin Popescu, Cecile Capponi, Stephane Ayache, Xavier BarĂ³

Authors must use the following template:

https://www.springer.com/gp/authors-editors/book-authors-editors/manuscript-preparation/5636

TeX-based submission preferred. Length requirements 15-30 pages. Submission portal: https://cmt3.research.microsoft.com/CIMLForecasting2018 (if you do not have a CMT account, please create one first and revisit the link).

Papers will be reviewed for scientific quality, notification date: April 15th 2019. The volume will appear in the Springer Series on Challenges in Machine Learning https://link.springer.com/bookseries/15602 Series Editors: Sergio Escalera, Isabelle Guyon, Hugo Jair Escalante

Topics of interest

This volume builds on preparatory work (including a workshop on Spatiotemporal Forecasting at NIPS 2016) culminating in the Big Data Technologies Horizon Prize of the European Commission. It aims to endow the discipline of forecasting, traditionally focused on applied mathematics and statistics with the end-to-end tools afforded by modern data science and machine learning: big spatiotemporal data and timely acquisition, uses of open-source analytics packages, deep learning techniques, general computing on graphics processor units, easy scalable deployment on virtualization services, uniform performance assessment of forecasting algorithms, customizable benchmarks both in terms of efficient algorithms and useful benchmark data sets.

Several important challenges were addressed by organizers and participants of the Big Data Horizon Prize competition. In its data science aspect, Big (Open) Data, while it provides a rich source of the information for the spatial/multivariate component of forecasting, it does not come with assurances of quality and availability, in archive or future form. Despite fully automatic code evaluation on hidden data, participants were allowed a wide choice of languages to submit their work in, support for a wide variety of analytics libraries (including GPU-enabled deep learning) was provided: managing software complexity was part of the challenge. In its statistics and learning theory aspect, forecasting inevitably involves the approximation of mappings that are highly stochastic and time-variable. Finally, Machine Learning systems that address big, nonstationary data processes must control over-parametrization and over-fit, and execute necessary steps (training, adaptation and prediction) within stringent time bounds, which are especially relevant in forecasting.

Submissions are welcome that address specific issues such as:

Methods/datasets:

  • Novel methods related to spatiotemporal forecasting: image parametrization and analysis such as dimensionality reduction, time series prediction methods over large state spaces, feature selection for forecasting, causal feature selection in multidimensional time series.
  • Novel methods to deal with wide data bandwidths in forecasting.
  • Novel methods for adaptive, quick turnaround deep learning in forecasting.
  • Hybrid models (gray-box models combining physics and sociology based methods with statistical methods, online learning, ensemble methods).
  • Benchmarks for forecasting video sequences.
  • Benchmarks for forecasting of time series.
  • Incremental learning and flexible resource-constrained learning.
  • Automatic machine learning for forecasting.
  • spatiotemporal datasets for forecasting in agriculture, epidemiology, geosciences, economics, animate motion.

Theory

  • Statistical comparison of quality in forecasting systems.
  • Stochastic forecasting evaluation.
  • Combinations of loss types for increased forecasting reliability.
  • Robust tail-estimation and distribution representation of stochastic forecasts.

Implementation:

  • Computational languages, platforms and toolboxes for forecasting.
  • Adaptation of computer vision toolboxes for video understanding and forecasting.
  • Big open data pipelines for spatiotemporal forecasting.

Applications, including:

  • Energy sector, grid, renewables.
  • Video-based applications, compression, predictive control systems using video streams.
  • Stochastic predictions in geosciences and meteorology using ML.
  • Epidemiology.
  • Economics.
  • Agriculture.
  • Multi-disciplinary and multi-modal convergence for event and trajectory forecasting.