Skip to search formSkip to main contentSkip to account menu
- Corpus ID: 269982742
@inproceedings{Shang2024GIFTUF, title={GIFT: Unlocking Full Potential of Labels in Distilled Dataset at Near-zero Cost}, author={Xinyi Shang and Peng Sun and Tao Lin}, year={2024}, url={https://api.semanticscholar.org/CorpusID:269982742}}
- Xinyi Shang, Peng Sun, Tao Lin
- Published 23 May 2024
- Computer Science
This paper introduces an extremely simple yet surprisingly effective plug-and-play approach, GIFT, which encompasses soft label refinement and a cosine similarity-based loss function to efficiently leverage full label information in dataset distillation.
Figures and Tables from this paper
- figure 1
- table 1
- table 2
- figure 2
- table 3
- figure 3
- table 4
- table 5
- figure 5
- table 6
- figure 6
- table 7
- figure 7
- table 8
- table 9
- table 10
- table 11
- table 12
- table 13
Ask This Paper
BETA
AI-Powered
Ask This Paper
BETA
AI-Powered
Unknown Error
An unexpected error occurred. Please try again.
No Answer Found
Ask another question that can be answered by this paper or rephrase your question.
We are still processing this paper
Please try again later.
Question Answering Unavailable
Please try again later.
No Response
The server took too long to answer your question. You can either rephrase your question or wait until it is less busy.
AI-Generated
Thank you for your feedback!
We're sorry, something went wrong while submitting this feedback.
Thank you for your feedback!
We're sorry, something went wrong while submitting this feedback.
Supporting Statements
Our system tries to constrain to information found in this paper. Results quality may vary. Learn more about how we generate these answers.
Feedback?
47 References
- Peng SunBei ShiDaiwei YuTao Lin
- 2023
Computer Science
ArXiv
This work proposes RDED, a novel computationally-efficient yet effective data distillation paradigm, to enable both diversity and realism of the distilled data in large-scale real-world applications.
- 7
- Highly Influential[PDF]
- Justin CuiRuochen WangSi SiCho-Jui Hsieh
- 2023
Computer Science
ICML
This work proposes a procedure to exactly compute the unrolled gradient with constant memory complexity, which allows to scale MTT to ImageNet-1K seamlessly with ~6x reduction in memory footprint, and discovers that it is challenging for M TT to handle datasets with a large number of classes, and proposes a novel soft label assignment that drastically improves its convergence.
- 56
- Highly Influential[PDF]
- Ziyao GuoKai WangGeorge CazenavetteHui LiKaipeng ZhangYang You
- 2023
Computer Science
ArXiv
This work proposes to align the difficulty of the generated patterns with the size of the synthetic dataset, and successfully scale trajectory matching-based methods to larger synthetic datasets, achieving lossless dataset distillation for the first time.
- 14
- Highly Influential[PDF]
- Zeyuan YinEric P. XingZhiqiang Shen
- 2023
Computer Science
NeurIPS
The proposed dataset condensation framework demonstrates flexibility across diverse dataset scales and exhibits multiple advantages in terms of arbitrary resolutions of synthesized images, low training cost and memory consumption with high-resolution synthesis, and the ability to scale up to arbitrary evaluation network architectures.
- 20
- Highly Influential[PDF]
- Geoffrey E. HintonO. VinyalsJ. Dean
- 2015
Computer Science
ArXiv
This work shows that it can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model and introduces a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse.
- 15,711
- Highly Influential[PDF]
- sh*tong ShaoZeyuan YinMuxin ZhouXindong ZhangZhiqiang Shen
- 2023
Computer Science
ArXiv
G-VBSM is the first algorithm to obtain strong performance across both small-scale and large-scale datasets and surpasses all SOTA methods by margins of 3.9%, 6.5%, and 10.1%, respectively.
- 6
- Highly Influential[PDF]
- Bo ZhaoHakan Bilen
- 2023
Computer Science
2023 IEEE/CVF Winter Conference on Applications…
This work proposes a simple yet effective method that synthesizes condensed images by matching feature distributions of the synthetic and original training images in many sampled embedding spaces and significantly reduces the synthesis cost while achieving comparable or better performance.
- 154
- Highly Influential[PDF]
- Ganlong ZhaoGuanbin LiYipeng QinYizhou Yu
- 2023
Computer Science
2023 IEEE/CVF Conference on Computer Vision and…
This paper identifies two important shortcomings of naive distribution matching and addresses them with three novel techniques and outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models.
- 23 [PDF]
- George CazenavetteTongzhou WangA. TorralbaAlexei A. EfrosJun-Yan Zhu
- 2023
Computer Science
2023 IEEE/CVF Conference on Computer Vision and…
A new optimization algorithm is presented that distills a large number of images into a few intermediate feature vectors in the generative model's latent space, significantly improving cross-architecture generalization in all settings.
- 36 [PDF]
- Zhendong YangAiling ZengZhe LiTianke ZhangChun YuanYu Li
- 2023
Computer Science
2023 IEEE/CVF International Conference on…
This work unifies the formulations of the two tasks by decomposing and reorganizing the generic KD loss into a Normalized KD (NKD) loss and customized soft labels for both target class (image’s category) and non-target classes named Universal Self-KD (USKD).
- 18 [PDF]
...
...
Related Papers
Showing 1 through 3 of 0 Related Papers