TTS-GAN: A Transformer-Based Time-Series Generative Adversarial Network

Li, Xiaomin; Metsis, Vangelis; Wang, Huangyingrui; Ngu, Anne H. H.

doi:10.1007/978-3-031-09342-5_13

Cited by 77 publications

(41 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The data augmentation method is a process of artificially increasing the amount of data by generating new data points from existing data that does not require substantial training data, including synthetic minority oversampling technique (SMOTE) [ 41 ], transformers [ 42 ], auto-Encoder [ 43 ], generative adversarial network (GAN) [ 44 ]. We have started experimentation with GAN for time series data in [ 45 ]. Recently, we have also used a GAN product from Gretel ( , accessed on 25 December 2022) to generate synthetic data.…”

Section: Discussionmentioning

confidence: 99%

Transfer Learning on Small Datasets for Improved Fall Detection

Maray

Ngu

et al. 2023

Sensors

View full text Add to dashboard Cite

Falls in the elderly are associated with significant morbidity and mortality. While numerous fall detection devices incorporating AI and machine learning algorithms have been developed, no known smartwatch-based system has been used successfully in real-time to detect falls for elderly persons. We have developed and deployed a SmartFall system on a commodity-based smartwatch which has been trialled by nine elderly participants. The system, while being usable and welcomed by the participants in our trials, has two serious limitations. The first limitation is the inability to collect a large amount of personalized data for training. When the fall detection model, which is trained with insufficient data, is used in the real world, it generates a large amount of false positives. The second limitation is the model drift problem. This means an accurate model trained using data collected with a specific device performs sub-par when used in another device. Therefore, building one model for each type of device/watch is not a scalable approach for developing smartwatch-based fall detection system. To tackle those issues, we first collected three datasets including accelerometer data for fall detection problem from different devices: the Microsoft watch (MSBAND), the Huawei watch, and the meta-sensor device. After that, a transfer learning strategy was applied to first explore the use of transfer learning to overcome the small dataset training problem for fall detection. We also demonstrated the use of transfer learning to generalize the model across the heterogeneous devices. Our preliminary experiments demonstrate the effectiveness of transfer learning for improving fall detection, achieving an F1 score higher by over 10% on average, an AUC higher by over 0.15 on average, and a smaller false positive prediction rate than the non-transfer learning approach across various datasets collected using different devices with different hardware specifications.

show abstract

Section: Discussionmentioning

confidence: 99%

Transfer Learning on Small Datasets for Improved Fall Detection

Maray

Ngu

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Further, Transformer-based GANs [73], [74] were proposed and showed the robustness of training. On the other hand, there are a few works that considered generating sequences using GAN, including the RNN with reinforcement training [75], [76], and the Transformerbased generator for text generation [77], [78], even long sequence generation [79], [80].…”

Section: B Generative Adversarial Networkmentioning

confidence: 99%

Generative Adversarial Networks for DNA Storage Channel Simulator

et al. 2023

View full text Add to dashboard Cite

DNA data storage systems have rapidly developed with novel error-correcting techniques, random access algorithms, and query systems. However, designing an algorithm for DNA storage systems is challenging, mainly due to the unpredictable nature of errors and the extremely high price of experiments. Thus, a simulator is of interest that can imitate the error statistics of a DNA storage system and replace the experiments in developing processes. We introduce novel generative adversarial networks that learn DNA storage channel statistics. Our simulator takes oligos (DNA sequences to write) as an input and generates a FASTQ file that includes output DNA reads and quality scores as if the oligos are synthesized and sequenced. We trained the proposed simulator with data from a single experiment consisting of 14,400 input oligo strands and 12,108,573 output reads. The error statistics between the input and the output of the trained generator match the actual error statistics, including the error rate at each position, the number of errors for each nucleotide, and high-order statistics. The code is available at https://github.com/gyfbianhuanyun/DNA_storage_simulator_GAN. INDEX TERMS channel simulator, DNA storage, generative adversarial networks, recurrent neural networks, transformer.

show abstract

“…TSG models have been developed to overcome insufficient training data due, for example, to acquisition difficulties or strict privacy constraints (Li et al, 2015;Esteban et al, 2017;Yoon et al, 2019;Ni et al, 2020a;Smith and Smith, 2020;Li et al, 2022;Zha, 2022). Current mainstream TSG methods consist of combining RNNs with the GAN architecture, as done in RCGAN (Esteban et al, 2017), TimeGAN (Yoon et al, 2019), and SigCWGAN (Ni et al, 2020a).…”

Section: Introductionmentioning

confidence: 99%

“…However, such methods are unable to effectively produce long sequences, given the limitation of the RNN model in keeping track of temporal dependencies between inputs that are distant in time (Vaswani et al, 2017). This problem is tackled by Li et al (2022) with TTS-CGAN (Transformer Time-Series Conditional GAN), where the RNN model is replaced by a transformer model (Vaswani et al, 2017). The challenges of training GANs, such as non-convergence, mode collapse, generator-discriminator unbalance, and sensitive hyperparameter selection, however, still remain.…”

Section: Introductionmentioning

confidence: 99%

“…To evaluate generated samples fairly, robust evaluation metrics and diverse benchmark datasets are necessary. There has, however, been a lack of agreement for which evaluation metrics to use in the TSG literature (Brophy et al, 2021) and most TSG studies have been evaluated only on a small range of datasets (Esteban et al, 2017;Yoon et al, 2019;Ni et al, 2020a;Li et al, 2022;Zha, 2022;Desai et al, 2021). A common visual comparative evaluation of TSG is carried out using PCA and t-SNE on time series (Yoon et al, 2019;Desai et al, 2021;Zha, 2022;Li et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Vector Quantized Time Series Generation with a Bidirectional Prior Model

Lee¹,

Malacarne²,

Aune³

2023

Preprint

View full text Add to dashboard Cite

Time series generation (TSG) studies have mainly focused on the use of Generative Adversarial Networks (GANs) combined with recurrent neural network (RNN) variants. However, the fundamental limitations and challenges of training GANs still remain. In addition, the RNN-family typically has difficulties with temporal consistency between distant timesteps. Motivated by the successes in the image generation (IMG) domain, we propose TimeVQVAE, the first work, to our knowledge, that uses vector quantization (VQ) techniques to address the TSG problem. Moreover, the priors of the discrete latent spaces are learned with bidirectional transformer models that can better capture global temporal consistency. We also propose VQ modeling in a time-frequency domain, separated into low-frequency (LF) and high-frequency (HF). This allows us to retain important characteristics of the time series and, in turn, generate new synthetic signals that are of better quality, with sharper changes in modularity, than its competing TSG methods. Our experimental evaluation is conducted on all datasets from the UCR archive, using well-established metrics in the IMG literature, such as Fréchet inception distance and inception scores. Our implementation on GitHub: https://github.com/ ML4ITS/TimeVQVAE.

show abstract

TTS-GAN: A Transformer-Based Time-Series Generative Adversarial Network

Cited by 77 publications

References 14 publications

Transfer Learning on Small Datasets for Improved Fall Detection

Transfer Learning on Small Datasets for Improved Fall Detection

Generative Adversarial Networks for DNA Storage Channel Simulator

Vector Quantized Time Series Generation with a Bidirectional Prior Model

Contact Info

Product

Resources

About