Pub Date : 2022-07-05DOI: 10.48550/arXiv.2207.01755
Yuhan Lin, Han Sun, Ningzhong Liu, Yetong Bian, Jun Cen, Huiyu Zhou
Due to the extreme complexity of scale and shape as well as the uncertainty of the predicted location, salient object detection in optical remote sensing images (RSI-SOD) is a very difficult task. The existing SOD methods can satisfy the detection performance for natural scene images, but they are not well adapted to RSI-SOD due to the above-mentioned image characteristics in remote sensing images. In this paper, we propose a novel Attention Guided Network (AGNet) for SOD in optical RSIs, including position enhancement stage and detail refinement stage. Specifically, the position enhancement stage consists of a semantic attention module and a contextual attention module to accurately describe the approximate location of salient objects. The detail refinement stage uses the proposed self-refinement module to progressively refine the predicted results under the guidance of attention and reverse attention. In addition, the hybrid loss is applied to supervise the training of the network, which can improve the performance of the model from three perspectives of pixel, region and statistics. Extensive experiments on two popular benchmarks demonstrate that AGNet achieves competitive performance compared to other state-of-the-art methods. The code will be available at https://github.com/NuaaYH/AGNet.
{"title":"Attention Guided Network for Salient Object Detection in Optical Remote Sensing Images","authors":"Yuhan Lin, Han Sun, Ningzhong Liu, Yetong Bian, Jun Cen, Huiyu Zhou","doi":"10.48550/arXiv.2207.01755","DOIUrl":"https://doi.org/10.48550/arXiv.2207.01755","url":null,"abstract":"Due to the extreme complexity of scale and shape as well as the uncertainty of the predicted location, salient object detection in optical remote sensing images (RSI-SOD) is a very difficult task. The existing SOD methods can satisfy the detection performance for natural scene images, but they are not well adapted to RSI-SOD due to the above-mentioned image characteristics in remote sensing images. In this paper, we propose a novel Attention Guided Network (AGNet) for SOD in optical RSIs, including position enhancement stage and detail refinement stage. Specifically, the position enhancement stage consists of a semantic attention module and a contextual attention module to accurately describe the approximate location of salient objects. The detail refinement stage uses the proposed self-refinement module to progressively refine the predicted results under the guidance of attention and reverse attention. In addition, the hybrid loss is applied to supervise the training of the network, which can improve the performance of the model from three perspectives of pixel, region and statistics. Extensive experiments on two popular benchmarks demonstrate that AGNet achieves competitive performance compared to other state-of-the-art methods. The code will be available at https://github.com/NuaaYH/AGNet.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"99 1","pages":"25-36"},"PeriodicalIF":0.0,"publicationDate":"2022-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81279505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-16DOI: 10.48550/arXiv.2206.08224
Panpan Zou, Yinglei Teng, Tao Niu
Online knowledge distillation conducts knowledge transfer among all student models to alleviate the reliance on pre-trained models. However, existing online methods rely heavily on the prediction distributions and neglect the further exploration of the representational knowledge. In this paper, we propose a novel Multi-scale Feature Extraction and Fusion method (MFEF) for online knowledge distillation, which comprises three key components: Multi-scale Feature Extraction, Dual-attention and Feature Fusion, towards generating more informative feature maps for distillation. The multiscale feature extraction exploiting divide-and-concatenate in channel dimension is proposed to improve the multi-scale representation ability of feature maps. To obtain more accurate information, we design a dual-attention to strengthen the important channel and spatial regions adaptively. Moreover, we aggregate and fuse the former processed feature maps via feature fusion to assist the training of student models. Extensive experiments on CIF AR-10, CIF AR-100, and CINIC-10 show that MFEF transfers more beneficial representational knowledge for distillation and outperforms alternative methods among various network architectures
{"title":"Multi scale Feature Extraction and Fusion for Online Knowledge Distillation","authors":"Panpan Zou, Yinglei Teng, Tao Niu","doi":"10.48550/arXiv.2206.08224","DOIUrl":"https://doi.org/10.48550/arXiv.2206.08224","url":null,"abstract":"Online knowledge distillation conducts knowledge transfer among all student models to alleviate the reliance on pre-trained models. However, existing online methods rely heavily on the prediction distributions and neglect the further exploration of the representational knowledge. In this paper, we propose a novel Multi-scale Feature Extraction and Fusion method (MFEF) for online knowledge distillation, which comprises three key components: Multi-scale Feature Extraction, Dual-attention and Feature Fusion, towards generating more informative feature maps for distillation. The multiscale feature extraction exploiting divide-and-concatenate in channel dimension is proposed to improve the multi-scale representation ability of feature maps. To obtain more accurate information, we design a dual-attention to strengthen the important channel and spatial regions adaptively. Moreover, we aggregate and fuse the former processed feature maps via feature fusion to assist the training of student models. Extensive experiments on CIF AR-10, CIF AR-100, and CINIC-10 show that MFEF transfers more beneficial representational knowledge for distillation and outperforms alternative methods among various network architectures","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"39 3-4 1","pages":"126-138"},"PeriodicalIF":0.0,"publicationDate":"2022-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77867222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-27DOI: 10.48550/arXiv.2205.13750
Jingjun Yi, Beichen Zhou
Multiple instance learning is qualified for many pattern recognition tasks with weakly annotated data. The combination of artificial neural network and multiple instance learning offers an end-to-end solution and has been widely utilized. However, challenges remain in two-folds. Firstly, current MIL pooling operators are usually pre-defined and lack flexibility to mine key instances. Secondly, in current solutions, the bag-level representation can be inaccurate or inaccessible. To this end, we propose an attention awareness multiple instance neural network framework in this paper. It consists of an instance-level classifier, a trainable MIL pooling operator based on spatial attention and a bag-level classification layer. Exhaustive experiments on a series of pattern recognition tasks demonstrate that our framework outperforms many state-of-the-art MIL methods and val-idates the effectiveness of our proposed attention MIL pooling operators.
{"title":"Attention Awareness Multiple Instance Neural Network","authors":"Jingjun Yi, Beichen Zhou","doi":"10.48550/arXiv.2205.13750","DOIUrl":"https://doi.org/10.48550/arXiv.2205.13750","url":null,"abstract":"Multiple instance learning is qualified for many pattern recognition tasks with weakly annotated data. The combination of artificial neural network and multiple instance learning offers an end-to-end solution and has been widely utilized. However, challenges remain in two-folds. Firstly, current MIL pooling operators are usually pre-defined and lack flexibility to mine key instances. Secondly, in current solutions, the bag-level representation can be inaccurate or inaccessible. To this end, we propose an attention awareness multiple instance neural network framework in this paper. It consists of an instance-level classifier, a trainable MIL pooling operator based on spatial attention and a bag-level classification layer. Exhaustive experiments on a series of pattern recognition tasks demonstrate that our framework outperforms many state-of-the-art MIL methods and val-idates the effectiveness of our proposed attention MIL pooling operators.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"127 1","pages":"581-592"},"PeriodicalIF":0.0,"publicationDate":"2022-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75815708","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-03DOI: 10.48550/arXiv.2205.01492
Thibaut Kulak, Anthony Fillion, Franccois Blayo
We propose a unified view on two widely used data visualization techniques: Self-Organizing Maps (SOMs) and Stochastic Neighbor Embedding (SNE). We show that they can both be derived from a common mathematical framework. Leveraging this formulation, we propose to compare SOM and SNE quantitatively on two datasets, and discuss possible avenues for future work to take advantage of both approaches.
{"title":"A unified view on Self-Organizing Maps (SOMs) and Stochastic Neighbor Embedding (SNE)","authors":"Thibaut Kulak, Anthony Fillion, Franccois Blayo","doi":"10.48550/arXiv.2205.01492","DOIUrl":"https://doi.org/10.48550/arXiv.2205.01492","url":null,"abstract":"We propose a unified view on two widely used data visualization techniques: Self-Organizing Maps (SOMs) and Stochastic Neighbor Embedding (SNE). We show that they can both be derived from a common mathematical framework. Leveraging this formulation, we propose to compare SOM and SNE quantitatively on two datasets, and discuss possible avenues for future work to take advantage of both approaches.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"139 20","pages":"458-468"},"PeriodicalIF":0.0,"publicationDate":"2022-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91403086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.48550/arXiv.2204.08726
Kenneth T. Co, David Martínez-Rego, Zhongyuan Hau, Emil C. Lupu
Deep neural networks have become an integral part of our software infrastructure and are being deployed in many widely-used and safety-critical applications. However, their integration into many systems also brings with it the vulnerability to test time attacks in the form of Universal Adversarial Perturbations (UAPs). UAPs are a class of perturbations that when applied to any input causes model misclassification. Although there is an ongoing effort to defend models against these adversarial attacks, it is often difficult to reconcile the trade-offs in model accuracy and robustness to adversarial attacks. Jacobian regularization has been shown to improve the robustness of models against UAPs, whilst model ensembles have been widely adopted to improve both predictive performance and model robustness. In this work, we propose a novel approach, Jacobian Ensembles-a combination of Jacobian regularization and model ensembles to significantly increase the robustness against UAPs whilst maintaining or improving model accuracy. Our results show that Jacobian Ensembles achieves previously unseen levels of accuracy and robustness, greatly improving over previous methods that tend to skew towards only either accuracy or robustness.
{"title":"Jacobian Ensembles Improve Robustness Trade-offs to Adversarial Attacks","authors":"Kenneth T. Co, David Martínez-Rego, Zhongyuan Hau, Emil C. Lupu","doi":"10.48550/arXiv.2204.08726","DOIUrl":"https://doi.org/10.48550/arXiv.2204.08726","url":null,"abstract":"Deep neural networks have become an integral part of our software infrastructure and are being deployed in many widely-used and safety-critical applications. However, their integration into many systems also brings with it the vulnerability to test time attacks in the form of Universal Adversarial Perturbations (UAPs). UAPs are a class of perturbations that when applied to any input causes model misclassification. Although there is an ongoing effort to defend models against these adversarial attacks, it is often difficult to reconcile the trade-offs in model accuracy and robustness to adversarial attacks. Jacobian regularization has been shown to improve the robustness of models against UAPs, whilst model ensembles have been widely adopted to improve both predictive performance and model robustness. In this work, we propose a novel approach, Jacobian Ensembles-a combination of Jacobian regularization and model ensembles to significantly increase the robustness against UAPs whilst maintaining or improving model accuracy. Our results show that Jacobian Ensembles achieves previously unseen levels of accuracy and robustness, greatly improving over previous methods that tend to skew towards only either accuracy or robustness.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"65 1","pages":"680-691"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79248862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-14DOI: 10.1007/978-3-031-15937-4_22
Andrea Castellani, Sebastian Schmitt, Barbara Hammer
{"title":"Stream-based Active Learning with Verification Latency in Non-stationary Environments","authors":"Andrea Castellani, Sebastian Schmitt, Barbara Hammer","doi":"10.1007/978-3-031-15937-4_22","DOIUrl":"https://doi.org/10.1007/978-3-031-15937-4_22","url":null,"abstract":"","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"6 2 1","pages":"260-272"},"PeriodicalIF":0.0,"publicationDate":"2022-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90498295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-13DOI: 10.48550/arXiv.2204.06337
Pervaiz Iqbal Khan, Imran Razzak, A. Dengel, Sheraz Ahmed
Health mention classification deals with the disease detection in a given text containing disease words. However, non-health and figurative use of disease words adds challenges to the task. Recently, adversarial training acting as a means of regularization has gained popularity in many NLP tasks. In this paper, we propose a novel approach to train language models for health mention classification of tweets that involves adversarial training. We generate adversarial examples by adding perturbation to the representations of transformer models for tweet examples at various levels using Gaussian noise. Further, we employ contrastive loss as an additional objective function. We evaluate the proposed method on the PHM2017 dataset extended version. Results show that our proposed approach improves the performance of classifier significantly over the baseline methods. Moreover, our analysis shows that adding noise at earlier layers improves models' performance whereas adding noise at intermediate layers deteriorates models' performance. Finally, adding noise towards the final layers performs better than the middle layers noise addition.
{"title":"A Novel Approach to Train Diverse Types of Language Models for Health Mention Classification of Tweets","authors":"Pervaiz Iqbal Khan, Imran Razzak, A. Dengel, Sheraz Ahmed","doi":"10.48550/arXiv.2204.06337","DOIUrl":"https://doi.org/10.48550/arXiv.2204.06337","url":null,"abstract":"Health mention classification deals with the disease detection in a given text containing disease words. However, non-health and figurative use of disease words adds challenges to the task. Recently, adversarial training acting as a means of regularization has gained popularity in many NLP tasks. In this paper, we propose a novel approach to train language models for health mention classification of tweets that involves adversarial training. We generate adversarial examples by adding perturbation to the representations of transformer models for tweet examples at various levels using Gaussian noise. Further, we employ contrastive loss as an additional objective function. We evaluate the proposed method on the PHM2017 dataset extended version. Results show that our proposed approach improves the performance of classifier significantly over the baseline methods. Moreover, our analysis shows that adding noise at earlier layers improves models' performance whereas adding noise at intermediate layers deteriorates models' performance. Finally, adding noise towards the final layers performs better than the middle layers noise addition.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"157 1","pages":"136-147"},"PeriodicalIF":0.0,"publicationDate":"2022-04-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73694540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-12DOI: 10.48550/arXiv.2204.05764
Stefan Pócos, Iveta Becková, I. Farkaš
Deep neural networks achieve remarkable performance in multiple fields. However, after proper training they suffer from an inherent vulnerability against adversarial examples (AEs). In this work we shed light on inner representations of the AEs by analysing their activations on the hidden layers. We test various types of AEs, each crafted using a specific norm constraint, which affects their visual appearance and eventually their behavior in the trained networks. Our results in image classification tasks (MNIST and CIFAR-10) reveal qualitative differences between the individual types of AEs, when comparing their proximity to the class-specific manifolds on the inner representations. We propose two methods that can be used to compare the distances to class-specific manifolds, regardless of the changing dimensions throughout the network. Using these methods, we consistently confirm that some of the adversarials do not necessarily leave the proximity of the manifold of the correct class, not even in the last hidden layer of the neural network. Next, using UMAP visualisation technique, we project the class activations to 2D space. The results indicate that the activations of the individual AEs are entangled with the activations of the test set. This, however, does not hold for a group of crafted inputs called the rubbish class. We also confirm the entanglement of adversarials with the test set numerically using the soft nearest neighbour loss.
{"title":"Examining the Proximity of Adversarial Examples to Class Manifolds in Deep Networks","authors":"Stefan Pócos, Iveta Becková, I. Farkaš","doi":"10.48550/arXiv.2204.05764","DOIUrl":"https://doi.org/10.48550/arXiv.2204.05764","url":null,"abstract":"Deep neural networks achieve remarkable performance in multiple fields. However, after proper training they suffer from an inherent vulnerability against adversarial examples (AEs). In this work we shed light on inner representations of the AEs by analysing their activations on the hidden layers. We test various types of AEs, each crafted using a specific norm constraint, which affects their visual appearance and eventually their behavior in the trained networks. Our results in image classification tasks (MNIST and CIFAR-10) reveal qualitative differences between the individual types of AEs, when comparing their proximity to the class-specific manifolds on the inner representations. We propose two methods that can be used to compare the distances to class-specific manifolds, regardless of the changing dimensions throughout the network. Using these methods, we consistently confirm that some of the adversarials do not necessarily leave the proximity of the manifold of the correct class, not even in the last hidden layer of the neural network. Next, using UMAP visualisation technique, we project the class activations to 2D space. The results indicate that the activations of the individual AEs are entangled with the activations of the test set. This, however, does not hold for a group of crafted inputs called the rubbish class. We also confirm the entanglement of adversarials with the test set numerically using the soft nearest neighbour loss.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"17 1","pages":"645-656"},"PeriodicalIF":0.0,"publicationDate":"2022-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74886642","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-11DOI: 10.48550/arXiv.2204.05077
Katsiaryna Haitsiukevich, A. Ilin
Modeling of conservative systems with neural networks is an area of active research. A popular approach is to use Hamiltonian neural networks (HNNs) which rely on the assumptions that a conservative system is described with Hamilton's equations of motion. Many recent works focus on improving the integration schemes used when training HNNs. In this work, we propose to enhance HNNs with an estimation of a continuous-time trajectory of the modeled system using an additional neural network, called a deep hidden physics model in the literature. We demonstrate that the proposed integration scheme works well for HNNs, especially with low sampling rates, noisy and irregular observations.
{"title":"Learning Trajectories of Hamiltonian Systems with Neural Networks","authors":"Katsiaryna Haitsiukevich, A. Ilin","doi":"10.48550/arXiv.2204.05077","DOIUrl":"https://doi.org/10.48550/arXiv.2204.05077","url":null,"abstract":"Modeling of conservative systems with neural networks is an area of active research. A popular approach is to use Hamiltonian neural networks (HNNs) which rely on the assumptions that a conservative system is described with Hamilton's equations of motion. Many recent works focus on improving the integration schemes used when training HNNs. In this work, we propose to enhance HNNs with an estimation of a continuous-time trajectory of the modeled system using an additional neural network, called a deep hidden physics model in the literature. We demonstrate that the proposed integration scheme works well for HNNs, especially with low sampling rates, noisy and irregular observations.","PeriodicalId":93416,"journal":{"name":"Artificial neural networks, ICANN : international conference ... proceedings. International Conference on Artificial Neural Networks (European Neural Network Society)","volume":"21 1","pages":"562-573"},"PeriodicalIF":0.0,"publicationDate":"2022-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77754708","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：