Imitation learning aims to reproduce expert behaviors without relying on an explicit reward signal. However, real-world demonstrations often present challenges, such as multi-modal, data imbalance, and expensive labeling processes. In this work, we propose a novel semi-supervised imitation learning architecture that learns disentangled behavior representations from imbalanced demonstrations using limited labeled data. Specifically, our method consists of three key components. First, we adapt the concept of semi-supervised generative adversarial networks to the imitation learning context. Second, we employ a learnable latent distribution to align the generated and expert data distributions. Finally, we utilize a regularized information maximization approach in conjunction with an approximate label prior to further improve the semi-supervised learning performance. Experimental results demonstrate the efficiency of our method in learning multi-modal behaviors from imbalanced demonstrations compared to baseline methods.

I ( x; y; c) can be decomposed as I (x; y; c) = I ( y; x) + I ( c; x) I ( y, c; x) = I ( y; x) + I ( c; x) H (y, c) + H (y, c |x) = I ( y; c) I (y; c |x). I ( s, a; s, a) is finally increased as well. The main parameters for training Ess-InfoGAIL are listed in Table 4. To minimize computational time, we restrict the update of the latent skill distribution to only the first iteration of policy updates. Our experiments demonstrate that this approach does not result in significant performance degradation.

Imitation learning aims to reproduce expert behaviors without relying on an explicit reward signal. However, real-world demonstrations often present challenges, such as multi-modal, data imbalance, and expensive labeling processes. In this work, we propose a novel semi-supervised imitation learning architecture that learns disentangled behavior representations from imbalanced demonstrations using limited labeled data. Specifically, our method consists of three key components. First, we adapt the concept of semi-supervised generative adversarial networks to the imitation learning context. Second, we employ a learnable latent distribution to align the generated and expert data distributions.