FelixOtt/gnss_class_incremental_learning

Bayesian Learning-driven Class-Incremental Learning for Adaptive GNSS Interference Classification

Latest update: 2025-04-07

Project size: 511 KB

Introduction

We propose a prototypical network with a Bayesian learning-driven contrastive loss (BLCL) tailored specifically for class-incremental learning scenarios. Therefore, we introduce a contrastive loss that incorporates new classes into the latent representation by reducing the intra-class distance and increasing the inter-class distance. Our approach dynamically adapts the balance between the cross-entropy and contrastive loss functions with a Bayesian learning technique. Empirical evaluations conducted on both the CIFAR-10 dataset for image classification and images of a GNSS-based dataset for interference classification validate the efficacy of our method, showcasing its superiority over existing state-of-the-art approaches.

Methods Reproduced and Included in this Repository

The following methods can be reproduced:

• BLCl: Bayesian learning-driven prototypical contrastive loss for class-incremental learning
• FineTune: Baseline method which simply updates parameters on new tasks
• EWC: Overcoming catastrophic forgetting in neural networks
• LwF: Learning without forgetting
• Replay: Baseline method with exemplar replay
• iCaRL: Incremental classifier and representation learning
• MEMO: A model or 603 exemplars: towards memory-efficient class-incremental learning

Dependencies

1. torch 1.81
2. torchvision 0.6.0
3. tqdm
4. numpy
5. scipy
6. Tensorboard

Run experiment

1. Run: python main.py --config=./exps/blcl.json --dataset=cifar10 --sumarry_file=CIFAR10_BLCL8_0123 --layers=[[2,2],[2,1],[1,1],[2]] --type=normal --convnet_type=memo_resnet56

2. Edit the [MODEL NAME].json file for global settings.

3. Edit the hyperparameters in the corresponding [MODEL NAME].py file (e.g., models/blcl.py).

4. --layers=BLCL specialized blocks (BLCL_[0123] corresponds to [[2,2],[2,1],[1,1],[2]])

5. --dataset=GNSS or --dataset=cifar10

6. --type=... whether to average the specialized blocks (avg if yes, normal" if no)

7. --convnet_type=... Type of ResNet block: memo_resnet18, memo_resnet32, memo_resnet44, memo_resnet56

Datasets

We have implemented the pre-processing of GNSS and CIFAR10. When training on CIFAR10, this framework will automatically download it.

For our spectrum highway dataset 1 recorded along a highway bridge, see the following website: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss/FIOT_highway

For our spectrum highway dataset 2 recorded along a highway bridge, see the following website: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss/fiot_highway2

For our dataset recorded in a controlled environment with a high-frequency antenna, see the following website: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss/FIOT_LC_laboratory

For our dataset recorded in a controlled environment with a low-frequency antenna, see the following website: https://gitlab.cc-asp.fraunhofer.de/darcy_gnss/controlled_low_frequency

Evaluation

1. cluster_analysis.py: Performs cluster analysis using embeddings and ground truth labels. Calculates the Davies-Bouldin and Calinski-Harabasz scores.

2. tsne.py

   • plots the embeddings as tsne plots
   • run: python tsne.py (datset name) (task), e.g., python tsne.py cifar 0

References

For more information of the dataset and results, see our publication. If you use our dataset or method for your research, please consider citing:

@inproceedings{raichur_heublein,
  author = {Nisha L. Raichur and Lucas Heublein and Tobias Feigl and Alexander Rügamer and Christopher Mutschler and Felix Ott},
  title = {{Bayesian Learning-driven Prototypical Contrastive Loss for Class-Incremental Learning}},
  booktitle = {\href{https://openreview.net/forum?id=dNWaTuKV9M}{Trans. on Machine Learning Research (TMLR)}},
  month = apr,
  year = {2025}
}

For further related publications, see:

@inproceedings{ott_heublein_icl,
  author = {Felix Ott and Lucas Heublein and Nisha Lakshmana Raichur and Tobias Feigl and Jonathan Hansen and Alexander Rügamer and Christopher Mutschler},
  title = {{Few-Shot Learning with Uncertainty-based Quadruplet Selection for Interference Classification in GNSS Data}},
  booktitle = {\href{https://ieeexplore.ieee.org/document/10578525}{IEEE Intl. Conf. on Localization and GNSS (ICL-GNSS)}},
  month = jun,
  year = {2024},
  address = {Antwerp, Belgium},
  doi = {10.1109/ICL-GNSS60721.2024.10578525}
}

Acknowledgment

This work has been carried out within the DARCII project, funding code 50NA2401, supported by the German Federal Ministry for Economic Affairs and Climate Action (BMWK), managed by the German Space Agency at DLR and assisted by the Bundesnetzagentur (BNetzA) and the Federal Agency for Cartography and Geodesy (BKG). Additionally, this work was supported by the Bavarian Ministry of Economic Affairs, Regional Development and Energy through the Center for Analytics – Data – Applications (ADA-Center) within the framework of „BAYERN DIGITAL II“ (20-3410-2-9-8).

License

This work is licensed under a CC BY-NC-SA 4.0: Creative Commons Attribution-Noncommercial-ShareAlike, see https://creativecommons.org/licenses/by-nc-sa/4.0/

Kontakt

If you have any questions or tips to improve the datasets, contact us:

Felix Ott: felix.ott@iis.fraunhofer.de

Nisha Lakshmana Raichur: nisha.lakshmana.raichur@iis.fraunhofer.de

Nordostpark 84, 90411 Nürnberg, Germany, GoogleMaps

Fraunhofer Institute for Integrated Circuits IIS, Nürnberg, Germany