Using time-series privileged information for provably efficient learning of prediction models

Rickard K.A. Karlsson, Martin Willbo, Zeshan M. Hussain, Rahul G. Krishnan, David Sontag, Fredrik Johansson
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:5459-5484, 2022.

Abstract

We study prediction of future outcomes with supervised models that use privileged information during learning. The privileged information comprises samples of time series observed between the baseline time of prediction and the future outcome; this information is only available at training time which differs from the traditional supervised learning. Our question is when using this privileged data leads to more sample-efficient learning of models that use only baseline data for predictions at test time. We give an algorithm for this setting and prove that when the time series are drawn from a non-stationary Gaussian-linear dynamical system of fixed horizon, learning with privileged information is more efficient than learning without it. On synthetic data, we test the limits of our algorithm and theory, both when our assumptions hold and when they are violated. On three diverse real-world datasets, we show that our approach is generally preferable to classical learning, particularly when data is scarce. Finally, we relate our estimator to a distillation approach both theoretically and empirically.

Cite this Paper

BibTeX
@InProceedings{pmlr-v151-k-a-karlsson22a, title = { Using time-series privileged information for provably efficient learning of prediction models }, author = {K.A. Karlsson, Rickard and Willbo, Martin and Hussain, Zeshan M. and Krishnan, Rahul G. and Sontag, David and Johansson, Fredrik}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {5459--5484}, year = {2022}, editor = {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel}, volume = {151}, series = {Proceedings of Machine Learning Research}, month = {28--30 Mar}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v151/k-a-karlsson22a/k-a-karlsson22a.pdf}, url = {https://proceedings.mlr.press/v151/k-a-karlsson22a.html}, abstract = { We study prediction of future outcomes with supervised models that use privileged information during learning. The privileged information comprises samples of time series observed between the baseline time of prediction and the future outcome; this information is only available at training time which differs from the traditional supervised learning. Our question is when using this privileged data leads to more sample-efficient learning of models that use only baseline data for predictions at test time. We give an algorithm for this setting and prove that when the time series are drawn from a non-stationary Gaussian-linear dynamical system of fixed horizon, learning with privileged information is more efficient than learning without it. On synthetic data, we test the limits of our algorithm and theory, both when our assumptions hold and when they are violated. On three diverse real-world datasets, we show that our approach is generally preferable to classical learning, particularly when data is scarce. Finally, we relate our estimator to a distillation approach both theoretically and empirically. } }
Endnote
%0 Conference Paper %T Using time-series privileged information for provably efficient learning of prediction models %A Rickard K.A. Karlsson %A Martin Willbo %A Zeshan M. Hussain %A Rahul G. Krishnan %A David Sontag %A Fredrik Johansson %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-k-a-karlsson22a %I PMLR %P 5459--5484 %U https://proceedings.mlr.press/v151/k-a-karlsson22a.html %V 151 %X We study prediction of future outcomes with supervised models that use privileged information during learning. The privileged information comprises samples of time series observed between the baseline time of prediction and the future outcome; this information is only available at training time which differs from the traditional supervised learning. Our question is when using this privileged data leads to more sample-efficient learning of models that use only baseline data for predictions at test time. We give an algorithm for this setting and prove that when the time series are drawn from a non-stationary Gaussian-linear dynamical system of fixed horizon, learning with privileged information is more efficient than learning without it. On synthetic data, we test the limits of our algorithm and theory, both when our assumptions hold and when they are violated. On three diverse real-world datasets, we show that our approach is generally preferable to classical learning, particularly when data is scarce. Finally, we relate our estimator to a distillation approach both theoretically and empirically.
APA
K.A. Karlsson, R., Willbo, M., Hussain, Z.M., Krishnan, R.G., Sontag, D. & Johansson, F.. (2022). Using time-series privileged information for provably efficient learning of prediction models . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:5459-5484 Available from https://proceedings.mlr.press/v151/k-a-karlsson22a.html.

Related Material