%0 Generic %1 bengio2019metatransfer %A Bengio, Yoshua %A Deleu, Tristan %A Rahaman, Nasim %A Ke, Rosemary %A Lachapelle, Sébastien %A Bilaniuk, Olexa %A Goyal, Anirudh %A Pal, Christopher %D 2019 %K meta-learning %T A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms %U http://arxiv.org/abs/1901.10912 %X We propose to meta-learn causal structures based on how fast a learner adapts to new distributions arising from sparse distributional changes, e.g. due to interventions, actions of agents and other sources of non-stationarities. We show that under this assumption, the correct causal structural choices lead to faster adaptation to modified distributions because the changes are concentrated in one or just a few mechanisms when the learned knowledge is modularized appropriately. This leads to sparse expected gradients and a lower effective number of degrees of freedom needing to be relearned while adapting to the change. It motivates using the speed of adaptation to a modified distribution as a meta-learning objective. We demonstrate how this can be used to determine the cause-effect relationship between two observed variables. The distributional changes do not need to correspond to standard interventions (clamping a variable), and the learner has no direct knowledge of these interventions. We show that causal structures can be parameterized via continuous variables and learned end-to-end. We then explore how these ideas could be used to also learn an encoder that would map low-level observed variables to unobserved causal variables leading to faster adaptation out-of-distribution, learning a representation space where one can satisfy the assumptions of independent mechanisms and of small and sparse changes in these mechanisms due to actions and non-stationarities.

@misc{bengio2019metatransfer, abstract = {We propose to meta-learn causal structures based on how fast a learner adapts to new distributions arising from sparse distributional changes, e.g. due to interventions, actions of agents and other sources of non-stationarities. We show that under this assumption, the correct causal structural choices lead to faster adaptation to modified distributions because the changes are concentrated in one or just a few mechanisms when the learned knowledge is modularized appropriately. This leads to sparse expected gradients and a lower effective number of degrees of freedom needing to be relearned while adapting to the change. It motivates using the speed of adaptation to a modified distribution as a meta-learning objective. We demonstrate how this can be used to determine the cause-effect relationship between two observed variables. The distributional changes do not need to correspond to standard interventions (clamping a variable), and the learner has no direct knowledge of these interventions. We show that causal structures can be parameterized via continuous variables and learned end-to-end. We then explore how these ideas could be used to also learn an encoder that would map low-level observed variables to unobserved causal variables leading to faster adaptation out-of-distribution, learning a representation space where one can satisfy the assumptions of independent mechanisms and of small and sparse changes in these mechanisms due to actions and non-stationarities.}, added-at = {2019-03-26T19:37:56.000+0100}, author = {Bengio, Yoshua and Deleu, Tristan and Rahaman, Nasim and Ke, Rosemary and Lachapelle, Sébastien and Bilaniuk, Olexa and Goyal, Anirudh and Pal, Christopher}, biburl = {https://www.bibsonomy.org/bibtex/20df43b7875b8d296606413e382e0b35b/acercyc}, description = {A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms}, interhash = {2a5f716d66b6fb93255942a52f538022}, intrahash = {0df43b7875b8d296606413e382e0b35b}, keywords = {meta-learning}, note = {cite arxiv:1901.10912}, timestamp = {2019-03-26T19:37:56.000+0100}, title = {A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms}, url = {http://arxiv.org/abs/1901.10912}, year = 2019 }