• Psychology, Computer Science, Mathematics
  • Published in ArXiv 2019

On Value Discrepancy of Imitation Learning

@article{Xu2019OnVD,
  title={On Value Discrepancy of Imitation Learning},
  author={Tian Xu and Ziniu Li and Yang Yu},
  journal={ArXiv},
  year={2019},
  volume={abs/1911.07027}
}
Imitation learning trains a policy from expert demonstrations. Imitation learning approaches have been designed from various principles, such as behavioral cloning via supervised learning, apprenticeship learning via inverse reinforcement learning, and GAIL via generative adversarial learning. In this paper, we propose a framework to analyze the theoretical property of imitation learning approaches based on discrepancy propagation analysis. Under the infinite-horizon setting, the framework… CONTINUE READING

Figures, Tables, and Topics from this paper.

References

Publications referenced by this paper.
SHOWING 1-10 OF 26 REFERENCES

Generalization and equilibrium in generative adversarial nets (GANs) (invited talk)

VIEW 6 EXCERPTS
HIGHLY INFLUENTIAL

On the Discrimination-Generalization Tradeoff in GANs

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

Generative Adversarial Imitation Learning

VIEW 17 EXCERPTS
HIGHLY INFLUENTIAL

A Reduction from Apprenticeship Learning to Classification

VIEW 6 EXCERPTS
HIGHLY INFLUENTIAL

Apprenticeship learning using linear programming

VIEW 18 EXCERPTS
HIGHLY INFLUENTIAL

Apprenticeship learning via inverse reinforcement learning

VIEW 18 EXCERPTS
HIGHLY INFLUENTIAL

Efficient Training of Artificial Neural Networks for Autonomous Navigation

  • Dean Pomerleau
  • Mathematics, Medicine, Computer Science
  • Neural Computation
  • 1991
VIEW 9 EXCERPTS
HIGHLY INFLUENTIAL

Model-Free Imitation Learning with Policy Optimization

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL