Convolutional Shrinkage Neural Networks Based Model-Agnostic Meta-Learning for Few-Shot Learning

He, Yunpeng; Zang, Chuanzhi; Zeng, Peng; Dong, Qingwei; Liu, Ding; Liu, Yuqi

doi:10.1007/s11063-022-10894-7

Cited by 9 publications

(4 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By adjusting the loss function, Sharp-MAML effectively alleviates the gradient vanishing problem, thereby improving the generalization ability of the model. He et al [20] proposed CSNNs, a novel convolutional shrinkage neural network architecture for few-shot learning. CSNNs employ a shrinkage module to effectively suppress noise, thereby improving the model's robustness to noise.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

TMNIO:Triplet merged network with involution operators for improved few‐shot image classification

Lulu,

Ranhui,

Shaojie

et al. 2024

IET Image Processing

View full text Add to dashboard Cite

Few‐shot learning enables machines to learn efficiently from limited labelled data. However, existing few‐shot learning methods may perform poorly when there is a lack of sufficient samples, and may encounter problems such as domain shift or overfitting when applied to new domains or tasks. To address the issues of poor fitting and insufficient generalization ability in new domains, a new method called triplet merged network with involution operators (TMNIO) is proposed. This method employs dual encoders that extract common and distinctive features from the prototype network, thereby enhancing the model's feature extraction capability. To further improve this ability, the traditional convolutional kernels are replaced with involution operators, which not only reduce the parameter count but also enlarge the receptive field to better extract local feature information. Additionally, this method employs a two‐stage training strategy, where triplet loss is used in the first stage to train the model and enhance its robustness and generalization ability. Extensive experiments on the miniImageNet, Omniglot, and Caltech‐UCSD Birds‐200 (CUB) datasets have shown that our proposed method achieved significant improvements in both training speed and accuracy, particularly on the miniImageNet dataset, where it achieved an outstanding 10% performance improvement.

show abstract

Section: Related Workmentioning

confidence: 99%

“…He et al. [20] proposed CSNNs, a novel convolutional shrinkage neural network architecture for few‐shot learning. CSNNs employ a shrinkage module to effectively suppress noise, thereby improving the model's robustness to noise.…”

Section: Related Workmentioning

confidence: 99%

TMNIO:Triplet merged network with involution operators for improved few‐shot image classification

Lulu,

Ranhui,

Shaojie

et al. 2024

IET Image Processing

View full text Add to dashboard Cite

show abstract

“…These extreme scenarios typically have much lower occurrence possibility than normal scenarios and thus have the problem of insufficient samples. This will aggravate the low sample utilization issue and negatively impact the agent's capability to learn the optimal dispatch policy when applying RL (He et al, 2023).…”

Section: Introductionmentioning

confidence: 99%

Power system data-driven dispatch using improved scenario generation considering time-series correlations

Li,

Huang,

Liang

et al. 2023

Front. Energy Res.

View full text Add to dashboard Cite

Reinforcement learning (RL) is recently studied for realizing fast and adaptive power system dispatch under the increasing penetration of renewable energy. RL has the limitation of relying on samples for agent training, and the application in power systems often faces the difficulty of insufficient scenario samples. So, scenario generation is of great importance for the application of RL. However, most of the existing scenario generation methods cannot handle time-series correlation, especially the correlation over long time scales, when generating the scenario. To address this issue, this paper proposes an RL-based dispatch method which can generate power system operational scenarios with time-series correlation for the agent’s training. First, a time-generative adversarial network (GAN)-based scenario generation model is constructed, which generates system operational scenarios with long- and short-time scale time-series correlations. Next, the “N-1” security is ensured by simulating “N-1” branch contingencies in the agent’s training. Finally, the model is trained in parallel in an actual power system environment, and its effectiveness is verified by comparisons against benchmark methods.

show abstract

“…Object detection is a fundamental task in computer vision, requiring the completion of localization and classification of targets in images [1]. One limitation of traditional object detection algorithms relies on large-scale annotated data, while this requirement cannot hold on in practice since collecting enough labelled data is expensive [1][2][3][4]. On the other hand, for rare targets, we may not have access to any training data.…”

Section: Introductionmentioning

confidence: 99%

Cross-domain Constrained Network for Zero-shot Object Detection

Shi

Tan

et al. 2022

Preprint

View full text Add to dashboard Cite

Zero-shot object detection (ZSD) aims to recognize and locate unseen objects without training samples. Despite the rapid progress in the ZSD technique, most existing algorithms learn a mapping relationship between visual features and semantic space based on seen classes to apply the unseen classes without considering unseen class possible connections in the inference. Besides, due to the heterogeneous characteristics (category attributes and distribution variations) in different domains, the network’s mapping relationship alone between visual and semantic data will lead to inference bias. To address the above issues, we propose a novel cross-domain constrained network for ZSD, named constraint-ZSD. Specifically, we built a target association strategy to associate seen instances to learn unseen classes, which guarantees that at least one unseen object is associated with a similar seen object during the training period. Then, we model a cross-domain constrained mechanism to set a relational knowledge graph and explicitly constrain the consistency of category correlation under visual and semantic domains. As a result, the detector is guided to learn object-related and domain-independent feature representations. Finally, extensive experiments are presented for two standard benchmarks of ZSD, PASCAL VOC and MSCOCO. Our method outperforms previous state-of-the-art techniques on ZSD and generalized ZSD (GZSD) tasks.

show abstract

Convolutional Shrinkage Neural Networks Based Model-Agnostic Meta-Learning for Few-Shot Learning

Cited by 9 publications

References 21 publications

TMNIO:Triplet merged network with involution operators for improved few‐shot image classification

TMNIO:Triplet merged network with involution operators for improved few‐shot image classification

Power system data-driven dispatch using improved scenario generation considering time-series correlations

Cross-domain Constrained Network for Zero-shot Object Detection

Contact Info

Product

Resources

About