End-to-End Semi-supervised Learning for Differentiable Particle Filters

Wen, Hao; Chen, Xiongjie; Papagiannis, Georgios; Hu, Conghui; Li, Yunpeng

doi:10.1109/icra48506.2021.9561889

Cited by 16 publications

(49 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The optimization of DPFs [7]- [9], [19] relies on the ground truth state information, and the objective function can be the rooted mean square error (RMSE) between the ground truth state and particle mean, or negative log-likelihood (NLL) of the ground truth state under the approximated posterior distribution. In the semi-supervised DPFs (SDPFs) [11], a pseudolikelihood function is maximized to utilize unlabeled state in parameter optimization. Specifically, by dividing observations, actions, and states into m blocks of length L, the pseudolikelihood used in [11] can be formulated as:…”

Section: B Differentiable Particle Filtersmentioning

confidence: 99%

“…In the semi-supervised DPFs (SDPFs) [11], a pseudolikelihood function is maximized to utilize unlabeled state in parameter optimization. Specifically, by dividing observations, actions, and states into m blocks of length L, the pseudolikelihood used in [11] can be formulated as:…”

Section: B Differentiable Particle Filtersmentioning

confidence: 99%

“…We also show that normalizing flows can be straightforwardly incorporated into dynamic models. We provide numerical implementation details for the proposed method based on the semi-supervised DPFs proposed in [11], which we name as the CNF-SDPF. To simplify the notation, we use θ to denote the set of parameters in the dynamic model, the measurement model, and the proposal distribution in this section.…”

Section: Conditional Normalizing Flow-based Differentiable Particle F...mentioning

confidence: 99%

“…The proposed conditional normalizing flow-based differentiable particle filter framework can be directly incorporated to different DPF variants. For illustration, we provide numerical implementation details of the CNF-DPF framework based on the semi-supervised differentiable particle filter (SDPF) [11] in Algorithm 2.…”

Section: Numerical Implementationmentioning

confidence: 99%

“…In [10], a fully differentiable resampling method was proposed based on the entropy-regularized optimal transport. A semi-supervised learning approach was introduced in [11] to optimize the parameters of DPFs by iteratively maximizing a pseudo-likelihood function so that observations without ground truth state information can be used in the optimization of model parameters. These variants of DPFs [6], [7], [11] are within the bootstrap particle filtering (BPF) framework [12], in which particles are proposed according to the dynamic model, which does not utilize the latest observation in generating the proposal distribution and can potentially lead to the weight degeneracy issue.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Differentiable Particle Filters through Conditional Normalizing Flow

Chen¹,

Wen²,

Liu³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Differentiable particle filters provide a flexible mechanism to adaptively train dynamic and measurement models by learning from observed data. However, most existing differentiable particle filters are within the bootstrap particle filtering framework and fail to incorporate the information from latest observations to construct better proposals. In this paper, we utilize conditional normalizing flows to construct proposal distributions for differentiable particle filters, enriching the distribution families that the proposal distributions can represent. In addition, normalizing flows are incorporated in the construction of the dynamic model, resulting in a more expressive dynamic model. We demonstrate the performance of the proposed conditional normalizing flow-based differentiable particle filters in a visual tracking task.

show abstract

Section: B Differentiable Particle Filtersmentioning

confidence: 99%

Section: B Differentiable Particle Filtersmentioning

confidence: 99%

Section: Conditional Normalizing Flow-based Differentiable Particle F...mentioning

confidence: 99%

Section: Numerical Implementationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Differentiable Particle Filters through Conditional Normalizing Flow

Chen¹,

Wen²,

Liu³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Enhancing State Estimation in Robots: A Data-Driven Approach with Differentiable Ensemble Kalman Filters

Liu,

Clark,

Campbell

et al. 2023

2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

Attention-Based End-to-End Differentiable Particle Filter for Audio Speaker Tracking

Zhao,

Xu,

Qian

et al. 2024

IEEE Open J. Signal Process.

View full text Add to dashboard Cite

Particle filters (PFs) have been widely used in speaker tracking due to their capability in modeling a non-linear process or a non-Gaussian environment. However, particle filters are limited by several issues. For example, pre-defined handcrafted measurements are often used which can limit the model performance. In addition, the transition and update models are often preset which make PF less flexible to be adapted to different scenarios. To address these issues, we propose an end-to-end differentiable particle filter framework by employing the multi-head attention to model the long-range dependencies. The proposed model employs the self-attention as the learned transition model and the cross-attention as the learned update model. To our knowledge, this is the first proposal of combining particle filter and transformer for speaker tracking, where the measurement extraction, transition and update steps are integrated into an end-to-end architecture. Experimental results show that the proposed model achieves superior performance over the recurrent baseline models.INDEX TERMS Particle filter, differentiable particle filter, transformer, end to end training, speaker tracking.

show abstract

End-to-End Semi-supervised Learning for Differentiable Particle Filters

Cited by 16 publications

References 19 publications

Differentiable Particle Filters through Conditional Normalizing Flow

Differentiable Particle Filters through Conditional Normalizing Flow

Enhancing State Estimation in Robots: A Data-Driven Approach with Differentiable Ensemble Kalman Filters

Attention-Based End-to-End Differentiable Particle Filter for Audio Speaker Tracking

Contact Info

Product

Resources

About