2023
|
Falkiewicz, Maciej; Takeishi, Naoya; Shekhzadeh, Imahn; Wehenkel, Antoine; Delaunoy, Arnaud; Louppe, Gilles; Kalousis, Alexandros Calibrating Neural Simulation-Based Inference with Differentiable Coverage Probability Inproceedings Advances in Neural Information Processing Systems 36, pp. 1082–1099, 2023. Abstract | Links | BibTeX @inproceedings{falkiewicz23,
title = {Calibrating Neural Simulation-Based Inference with Differentiable Coverage Probability},
author = {Maciej Falkiewicz and Naoya Takeishi and Imahn Shekhzadeh and Antoine Wehenkel and Arnaud Delaunoy and Gilles Louppe and Alexandros Kalousis},
url = {https://github.com/DMML-Geneva/calibrated-posterior},
doi = {10.48550/arXiv.2310.13402},
year = {2023},
date = {2023-10-20},
booktitle = {Advances in Neural Information Processing Systems 36},
pages = {1082--1099},
abstract = {Bayesian inference allows expressing the uncertainty of posterior belief under a probabilistic model given prior information and the likelihood of the evidence. Predominantly, the likelihood function is only implicitly established by a simulator posing the need for simulation-based inference (SBI). However, the existing algorithms can yield overconfident posteriors (Hermans et al., 2022) defeating the whole purpose of credibility if the uncertainty quantification is inaccurate. We propose to include a calibration term directly into the training objective of the neural model in selected amortized SBI techniques. By introducing a relaxation of the classical formulation of calibration error we enable end-to-end backpropagation. The proposed method is not tied to any particular neural model and brings moderate computational overhead compared to the profits it introduces. It is directly applicable to existing computational pipelines allowing reliable black-box posterior inference. We empirically show on six benchmark problems that the proposed method achieves competitive or better results in terms of coverage and expected posterior density than the previously existing approaches.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Bayesian inference allows expressing the uncertainty of posterior belief under a probabilistic model given prior information and the likelihood of the evidence. Predominantly, the likelihood function is only implicitly established by a simulator posing the need for simulation-based inference (SBI). However, the existing algorithms can yield overconfident posteriors (Hermans et al., 2022) defeating the whole purpose of credibility if the uncertainty quantification is inaccurate. We propose to include a calibration term directly into the training objective of the neural model in selected amortized SBI techniques. By introducing a relaxation of the classical formulation of calibration error we enable end-to-end backpropagation. The proposed method is not tied to any particular neural model and brings moderate computational overhead compared to the profits it introduces. It is directly applicable to existing computational pipelines allowing reliable black-box posterior inference. We empirically show on six benchmark problems that the proposed method achieves competitive or better results in terms of coverage and expected posterior density than the previously existing approaches. |
2022
|
Takeishi, Naoya; Kalousis, Alexandros Deep Grey-Box Modeling With Adaptive Data-Driven Models Toward Trustworthy Estimation of Theory-Driven Models Unpublished 2022, (arXiv:2210.13103). Abstract | Links | BibTeX @unpublished{TakeishiDeepGreyBox2022,
title = {Deep Grey-Box Modeling With Adaptive Data-Driven Models Toward Trustworthy Estimation of Theory-Driven Models},
author = {Naoya Takeishi and Alexandros Kalousis},
url = {https://arxiv.org/abs/2210.13103},
year = {2022},
date = {2022-10-24},
abstract = {The combination of deep neural nets and theory-driven models, which we call deep grey-box modeling, can be inherently interpretable to some extent thanks to the theory backbone. Deep grey-box models are usually learned with a regularized risk minimization to prevent a theory-driven part from being overwritten and ignored by a deep neural net. However, an estimation of the theory-driven part obtained by uncritically optimizing a regularizer can hardly be trustworthy when we are not sure what regularizer is suitable for the given data, which may harm the interpretability. Toward a trustworthy estimation of the theory-driven part, we should analyze regularizers' behavior to compare different candidates and to justify a specific choice. In this paper, we present a framework that enables us to analyze a regularizer's behavior empirically with a slight change in the neural net's architecture and the training objective.},
note = {arXiv:2210.13103},
keywords = {},
pubstate = {published},
tppubtype = {unpublished}
}
The combination of deep neural nets and theory-driven models, which we call deep grey-box modeling, can be inherently interpretable to some extent thanks to the theory backbone. Deep grey-box models are usually learned with a regularized risk minimization to prevent a theory-driven part from being overwritten and ignored by a deep neural net. However, an estimation of the theory-driven part obtained by uncritically optimizing a regularizer can hardly be trustworthy when we are not sure what regularizer is suitable for the given data, which may harm the interpretability. Toward a trustworthy estimation of the theory-driven part, we should analyze regularizers' behavior to compare different candidates and to justify a specific choice. In this paper, we present a framework that enables us to analyze a regularizer's behavior empirically with a slight change in the neural net's architecture and the training objective. |
2021
|
Takeishi, Naoya; Kalousis, Alexandros Variational Autoencoder with Differentiable Physics Engine for Human Gait Analysis and Synthesis Workshop Deep Generative Models and Downstream Applications Workshop, 2021. Abstract | Links | BibTeX @workshop{Takeishi2021b,
title = { Variational Autoencoder with Differentiable Physics Engine for Human Gait Analysis and Synthesis},
author = {Naoya Takeishi and Alexandros Kalousis},
url = {https://openreview.net/forum?id=9ISlKio3Bt},
year = {2021},
date = {2021-12-14},
booktitle = {Deep Generative Models and Downstream Applications Workshop},
abstract = {We address the task of learning generative models of human gait. As gait motion always follows the physical laws, a generative model should also produce outputs that comply with the physical laws, particularly rigid body dynamics with contact and friction. We propose a deep generative model combined with a differentiable physics engine, which outputs physically plausible signals by construction. The proposed model is also equipped with a policy network conditioned on each sample. We show an example of the application of such a model to style transfer of gait.},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
We address the task of learning generative models of human gait. As gait motion always follows the physical laws, a generative model should also produce outputs that comply with the physical laws, particularly rigid body dynamics with contact and friction. We propose a deep generative model combined with a differentiable physics engine, which outputs physically plausible signals by construction. The proposed model is also equipped with a policy network conditioned on each sample. We show an example of the application of such a model to style transfer of gait. |
Takeishi, Naoya; Kalousis, Alexandros Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling Inproceedings Advances in Neural Information Processing Systems 34, 2021. Abstract | Links | BibTeX @inproceedings{Takeishi2021a,
title = {Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling},
author = {Naoya Takeishi and Alexandros Kalousis},
url = {https://openreview.net/forum?id=0p0gt1Pn2Gv
https://github.com/n-takeishi/phys-vae},
year = {2021},
date = {2021-12-07},
booktitle = {Advances in Neural Information Processing Systems 34},
abstract = {Integrating physics models within machine learning models holds considerable promise toward learning robust models with improved interpretability and abilities to extrapolate. In this work, we focus on the integration of incomplete physics models into deep generative models. In particular, we introduce an architecture of variational autoencoders (VAEs) in which a part of the latent space is grounded by physics. A key technical challenge is to strike a balance between the incomplete physics and trainable components such as neural networks for ensuring that the physics part is used in a meaningful manner. To this end, we propose a regularized learning method that controls the effect of the trainable components and preserves the semantics of the physics-based latent variables as intended. We not only demonstrate generative performance improvements over a set of synthetic and real-world datasets, but we also show that we learn robust models that can consistently extrapolate beyond the training distribution in a meaningful manner. Moreover, we show that we can control the generative process in an interpretable manner.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Integrating physics models within machine learning models holds considerable promise toward learning robust models with improved interpretability and abilities to extrapolate. In this work, we focus on the integration of incomplete physics models into deep generative models. In particular, we introduce an architecture of variational autoencoders (VAEs) in which a part of the latent space is grounded by physics. A key technical challenge is to strike a balance between the incomplete physics and trainable components such as neural networks for ensuring that the physics part is used in a meaningful manner. To this end, we propose a regularized learning method that controls the effect of the trainable components and preserves the semantics of the physics-based latent variables as intended. We not only demonstrate generative performance improvements over a set of synthetic and real-world datasets, but we also show that we learn robust models that can consistently extrapolate beyond the training distribution in a meaningful manner. Moreover, we show that we can control the generative process in an interpretable manner. |