Traffic4cast at NeurIPS 2021 - Temporal and Spatial Few-Shot Transfer Learning in Gridded Geo-Spatial Processes

Christian Eichenberger, Moritz Neun, Henry Martin, Pedro Herruzo, Markus Spanring, Yichao Lu, Sungbin Choi, Vsevolod Konyakhin, Nina Lukashina, Aleksei Shpilman, Nina Wiedemann, Martin Raubal, Bo Wang, Hai L. Vu, Reza Mohajerpoor, Chen Cai, Inhi Kim, Luca Hermes, Andrew Melnik, Riza Velioglu, Markus Vieth, Malte Schilling, Alabi Bojesomo, Hasan Al Marzouqi, Panos Liatsis, Jay Santokhi, Dylan Hillier, Yiming Yang, Joned Sarwar, Anna Jordan, Emil Hewage, David Jonietz, Fei Tang, Aleksandra Gruca, Michael Kopp, David Kreil, Sepp Hochreiter
Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track, PMLR 176:97-112, 2022.

Abstract

The IARAI Traffic4cast competitions at NeurIPS 2019 and 2020 showed that neural networks can successfully predict future traffic conditions 1 hour into the future on simply aggregated GPS probe data in time and space bins. We thus reinterpreted the challenge of forecasting traffic conditions as a movie completion task. U-Nets proved to be the winning architecture, demonstrating an ability to extract relevant features in this complex real-world geo-spatial process. Building on the previous competitions, Traffic4cast 2021 now focuses on the question of model robustness and generalizability across time and space. Moving from one city to an entirely different city, or moving from pre-COVID times to times after COVID hit the world thus introduces a clear domain shift. We thus, for the first time, release data featuring such domain shifts. The competition now covers ten cities over 2 years, providing data compiled from over $10^{12}$ GPS probe data. Winning solutions captured traffic dynamics sufficiently well to even cope with these complex domain shifts. Surprisingly, this seemed to require only the previous 1h traffic dynamic history and static road graph as input.

Cite this Paper

BibTeX
@InProceedings{pmlr-v176-eichenberger22a, title = {Traffic4cast at NeurIPS 2021 - Temporal and Spatial Few-Shot Transfer Learning in Gridded Geo-Spatial Processes}, author = {Eichenberger, Christian and Neun, Moritz and Martin, Henry and Herruzo, Pedro and Spanring, Markus and Lu, Yichao and Choi, Sungbin and Konyakhin, Vsevolod and Lukashina, Nina and Shpilman, Aleksei and Wiedemann, Nina and Raubal, Martin and Wang, Bo and Vu, Hai L. and Mohajerpoor, Reza and Cai, Chen and Kim, Inhi and Hermes, Luca and Melnik, Andrew and Velioglu, Riza and Vieth, Markus and Schilling, Malte and Bojesomo, Alabi and Marzouqi, Hasan Al and Liatsis, Panos and Santokhi, Jay and Hillier, Dylan and Yang, Yiming and Sarwar, Joned and Jordan, Anna and Hewage, Emil and Jonietz, David and Tang, Fei and Gruca, Aleksandra and Kopp, Michael and Kreil, David and Hochreiter, Sepp}, booktitle = {Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track}, pages = {97--112}, year = {2022}, editor = {Kiela, Douwe and Ciccone, Marco and Caputo, Barbara}, volume = {176}, series = {Proceedings of Machine Learning Research}, month = {06--14 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v176/eichenberger22a/eichenberger22a.pdf}, url = {https://proceedings.mlr.press/v176/eichenberger22a.html}, abstract = {The IARAI Traffic4cast competitions at NeurIPS 2019 and 2020 showed that neural networks can successfully predict future traffic conditions 1 hour into the future on simply aggregated GPS probe data in time and space bins. We thus reinterpreted the challenge of forecasting traffic conditions as a movie completion task. U-Nets proved to be the winning architecture, demonstrating an ability to extract relevant features in this complex real-world geo-spatial process. Building on the previous competitions, Traffic4cast 2021 now focuses on the question of model robustness and generalizability across time and space. Moving from one city to an entirely different city, or moving from pre-COVID times to times after COVID hit the world thus introduces a clear domain shift. We thus, for the first time, release data featuring such domain shifts. The competition now covers ten cities over 2 years, providing data compiled from over $10^{12}$ GPS probe data. Winning solutions captured traffic dynamics sufficiently well to even cope with these complex domain shifts. Surprisingly, this seemed to require only the previous 1h traffic dynamic history and static road graph as input. } }
Endnote
%0 Conference Paper %T Traffic4cast at NeurIPS 2021 - Temporal and Spatial Few-Shot Transfer Learning in Gridded Geo-Spatial Processes %A Christian Eichenberger %A Moritz Neun %A Henry Martin %A Pedro Herruzo %A Markus Spanring %A Yichao Lu %A Sungbin Choi %A Vsevolod Konyakhin %A Nina Lukashina %A Aleksei Shpilman %A Nina Wiedemann %A Martin Raubal %A Bo Wang %A Hai L. Vu %A Reza Mohajerpoor %A Chen Cai %A Inhi Kim %A Luca Hermes %A Andrew Melnik %A Riza Velioglu %A Markus Vieth %A Malte Schilling %A Alabi Bojesomo %A Hasan Al Marzouqi %A Panos Liatsis %A Jay Santokhi %A Dylan Hillier %A Yiming Yang %A Joned Sarwar %A Anna Jordan %A Emil Hewage %A David Jonietz %A Fei Tang %A Aleksandra Gruca %A Michael Kopp %A David Kreil %A Sepp Hochreiter %B Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track %C Proceedings of Machine Learning Research %D 2022 %E Douwe Kiela %E Marco Ciccone %E Barbara Caputo %F pmlr-v176-eichenberger22a %I PMLR %P 97--112 %U https://proceedings.mlr.press/v176/eichenberger22a.html %V 176 %X The IARAI Traffic4cast competitions at NeurIPS 2019 and 2020 showed that neural networks can successfully predict future traffic conditions 1 hour into the future on simply aggregated GPS probe data in time and space bins. We thus reinterpreted the challenge of forecasting traffic conditions as a movie completion task. U-Nets proved to be the winning architecture, demonstrating an ability to extract relevant features in this complex real-world geo-spatial process. Building on the previous competitions, Traffic4cast 2021 now focuses on the question of model robustness and generalizability across time and space. Moving from one city to an entirely different city, or moving from pre-COVID times to times after COVID hit the world thus introduces a clear domain shift. We thus, for the first time, release data featuring such domain shifts. The competition now covers ten cities over 2 years, providing data compiled from over $10^{12}$ GPS probe data. Winning solutions captured traffic dynamics sufficiently well to even cope with these complex domain shifts. Surprisingly, this seemed to require only the previous 1h traffic dynamic history and static road graph as input.
APA
Eichenberger, C., Neun, M., Martin, H., Herruzo, P., Spanring, M., Lu, Y., Choi, S., Konyakhin, V., Lukashina, N., Shpilman, A., Wiedemann, N., Raubal, M., Wang, B., Vu, H.L., Mohajerpoor, R., Cai, C., Kim, I., Hermes, L., Melnik, A., Velioglu, R., Vieth, M., Schilling, M., Bojesomo, A., Marzouqi, H.A., Liatsis, P., Santokhi, J., Hillier, D., Yang, Y., Sarwar, J., Jordan, A., Hewage, E., Jonietz, D., Tang, F., Gruca, A., Kopp, M., Kreil, D. & Hochreiter, S.. (2022). Traffic4cast at NeurIPS 2021 - Temporal and Spatial Few-Shot Transfer Learning in Gridded Geo-Spatial Processes. Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track, in Proceedings of Machine Learning Research 176:97-112 Available from https://proceedings.mlr.press/v176/eichenberger22a.html.

Related Material