Capacitated Vehicle Routing Problem (CVRP) is a representative type of Vehicle Routing Problem (VRP) and it is NP-hard. With the increase of the scale of the problem, the existing method is easy to fall into a local optimal solution, and the solution time is too long. To overcome these problems, in this paper, we propose an Enhanced Adaptive Large Neighborhood Search algorithm (EALNS). The EALNS adds a new type of linear removal strategy and selects several adjacent nodes on a route to be removed so that the vehicle can serve more customers. In the ALNS decision-making stage, an adaptive mechanism that weighs the time factor is added, so that each strategy combination can adjust the weight according to the solved time. Experiments are performed through three internationally published benchmarks. Experimental results show that the EALNS is competitive and can obtain satisfactory results in most instances. We compare with the optimal results from the collective best results reported in the literature, EALNS improves 2.30% average accuracy and significantly reduces the average solution time.
{"title":"An Enhanced Adaptive Large Neighborhood Search Algorithm for the Capacitated Vehicle Routing Problem","authors":"Haiping Zhang, Wang Yang","doi":"10.1145/3457682.3457694","DOIUrl":"https://doi.org/10.1145/3457682.3457694","url":null,"abstract":"Capacitated Vehicle Routing Problem (CVRP) is a representative type of Vehicle Routing Problem (VRP) and it is NP-hard. With the increase of the scale of the problem, the existing method is easy to fall into a local optimal solution, and the solution time is too long. To overcome these problems, in this paper, we propose an Enhanced Adaptive Large Neighborhood Search algorithm (EALNS). The EALNS adds a new type of linear removal strategy and selects several adjacent nodes on a route to be removed so that the vehicle can serve more customers. In the ALNS decision-making stage, an adaptive mechanism that weighs the time factor is added, so that each strategy combination can adjust the weight according to the solved time. Experiments are performed through three internationally published benchmarks. Experimental results show that the EALNS is competitive and can obtain satisfactory results in most instances. We compare with the optimal results from the collective best results reported in the literature, EALNS improves 2.30% average accuracy and significantly reduces the average solution time.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133623194","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}
Yuhao Wu, Weipeng Cao, Ye Liu, Zhong Ming, Jian-qiang Li, Bo Lu
Zero-Shot Learning (ZSL) is an effective paradigm to solve label prediction when some classes have no training samples. In recent years, many ZSL algorithms have been proposed. Among them, semantic autoencoder (SAE) is widely used because of its simplicity and good generalization ability. However, our research found that most of the existing SAE based methods use implicit constraints to guarantee the mapping quality between feature space and semantic space. In fact, the implicit constraints are insufficient in minimizing the structural risk of the model and easy to cause the over-fitting problem. To solve this problem, we propose a novel SAE algorithm with the L2-norm constraint (SAE-L2) in this study. SAE-L2 adds the L2 regularization constraint to the mapping parameters in its optimization objective, which explicitly guarantees the structural risk minimization of the model. Extensive experiments on four benchmark datasets show that our proposed SAE-L2 can achieve better performance than the original SAE model and other ZSL algorithms.
{"title":"Semantic Auto-Encoder with L2-norm Constraint for Zero-Shot Learning","authors":"Yuhao Wu, Weipeng Cao, Ye Liu, Zhong Ming, Jian-qiang Li, Bo Lu","doi":"10.1145/3457682.3457699","DOIUrl":"https://doi.org/10.1145/3457682.3457699","url":null,"abstract":"Zero-Shot Learning (ZSL) is an effective paradigm to solve label prediction when some classes have no training samples. In recent years, many ZSL algorithms have been proposed. Among them, semantic autoencoder (SAE) is widely used because of its simplicity and good generalization ability. However, our research found that most of the existing SAE based methods use implicit constraints to guarantee the mapping quality between feature space and semantic space. In fact, the implicit constraints are insufficient in minimizing the structural risk of the model and easy to cause the over-fitting problem. To solve this problem, we propose a novel SAE algorithm with the L2-norm constraint (SAE-L2) in this study. SAE-L2 adds the L2 regularization constraint to the mapping parameters in its optimization objective, which explicitly guarantees the structural risk minimization of the model. Extensive experiments on four benchmark datasets show that our proposed SAE-L2 can achieve better performance than the original SAE model and other ZSL algorithms.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131552496","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}
With increased interests in Explainable Artificial Intelligence (XAI), many researches find ways to provide explanations for machine learning algorithms and their predictions. We propose Multiple Choice Questioned Convolutional Neural Network (MCQ-CNN) to better understand the prediction of image classifier by considering the problem of multi-class classification as the problem of multiple choice question. MCQ-CNN not only performs classification of the query image, but also explains the classification result by demonstrating the elimination process of multi-class labels in patch-level. The proposed model consists of two modules. Classification module is to classify class label of the query. Elimination module is to perform similarity measure in patch-level to distinguish whether the target object part shares the feature of certain class label or not. Classification module is trained using ResNet with high classification accuracy. Elimination module performs similarity measure by distance metric learning based on Large Margin Nearest Neighbor (LMNN). Our experiments have shown notable performances in both classification and elimination modules.
{"title":"Towards Explainable Image Classifier: An Analogy to Multiple Choice Question Using Patch-level Similarity Measure","authors":"Yian Seo, K. Shin","doi":"10.1145/3457682.3457730","DOIUrl":"https://doi.org/10.1145/3457682.3457730","url":null,"abstract":"With increased interests in Explainable Artificial Intelligence (XAI), many researches find ways to provide explanations for machine learning algorithms and their predictions. We propose Multiple Choice Questioned Convolutional Neural Network (MCQ-CNN) to better understand the prediction of image classifier by considering the problem of multi-class classification as the problem of multiple choice question. MCQ-CNN not only performs classification of the query image, but also explains the classification result by demonstrating the elimination process of multi-class labels in patch-level. The proposed model consists of two modules. Classification module is to classify class label of the query. Elimination module is to perform similarity measure in patch-level to distinguish whether the target object part shares the feature of certain class label or not. Classification module is trained using ResNet with high classification accuracy. Elimination module performs similarity measure by distance metric learning based on Large Margin Nearest Neighbor (LMNN). Our experiments have shown notable performances in both classification and elimination modules.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129944576","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}
Many creators find crowdfunding websites one of the best ways to get assistance for their campaigns. Kickstarter, as one representative crowdfunding website, provides a great platform for their brightest dreams. However, not everyone successfully reaches their funding goals. In this paper, we figure out what machine learning model and factors can best predict success probability in a Kickstarter campaign. Through comparing 6 different machine learning models, we find that the best performing model is the Random Forest model, with robust forecast accuracy of 87.85%, which is 10% higher than existing studies. Factor importance analysis indicates that the number of backers, whether picked up by editors, and the edit time of campaign are the top three most important factors in determining the success rate of crowd-funding projects. This suggests, to launch a successful project, the number of backers, whether picked up by editors, and the edit time of campaign should be weighted more than other factors. Our research shed light on both crowd-funding project determinants and machine leaning down-stream applications.
{"title":"Do You Want to Foresee Your Future? The Best Model Predicting the Success of Kickstarter Campaigns","authors":"Jiayu Tian","doi":"10.1145/3457682.3457716","DOIUrl":"https://doi.org/10.1145/3457682.3457716","url":null,"abstract":"Many creators find crowdfunding websites one of the best ways to get assistance for their campaigns. Kickstarter, as one representative crowdfunding website, provides a great platform for their brightest dreams. However, not everyone successfully reaches their funding goals. In this paper, we figure out what machine learning model and factors can best predict success probability in a Kickstarter campaign. Through comparing 6 different machine learning models, we find that the best performing model is the Random Forest model, with robust forecast accuracy of 87.85%, which is 10% higher than existing studies. Factor importance analysis indicates that the number of backers, whether picked up by editors, and the edit time of campaign are the top three most important factors in determining the success rate of crowd-funding projects. This suggests, to launch a successful project, the number of backers, whether picked up by editors, and the edit time of campaign should be weighted more than other factors. Our research shed light on both crowd-funding project determinants and machine leaning down-stream applications.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"130 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133087599","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}
Sha Wen, Kai Liu, Shaoqing Tian, Mingming Fan, Lin Yan
Previous infrared small target detection approaches mainly solve the problem of detecting small target in sky background with strong cloud occlusion. However, these methods hardly exclude the negative objects other than cloud that cause false alarms. To address this problem, we propose an infrared small target detection framework using segmentation based region proposal and Convolution Neural Network (SCNN). In specific, an improved segmentation algorithm is used to obtain the salient regions from the background as the proposals. To reduce the high false alarms from proposals, a lightweight CNN is used to classify these regions and make final predictions. Owning to the lack of current public infrared small target datasets, a new infrared dataset is proposed in this paper. The experimental results demonstrate that the proposed method has a good performance in detection rate and false alarm rate.
{"title":"An Infrared Small Target Detection Method Using Segmentation Based Region Proposal and CNN","authors":"Sha Wen, Kai Liu, Shaoqing Tian, Mingming Fan, Lin Yan","doi":"10.1145/3457682.3457705","DOIUrl":"https://doi.org/10.1145/3457682.3457705","url":null,"abstract":"Previous infrared small target detection approaches mainly solve the problem of detecting small target in sky background with strong cloud occlusion. However, these methods hardly exclude the negative objects other than cloud that cause false alarms. To address this problem, we propose an infrared small target detection framework using segmentation based region proposal and Convolution Neural Network (SCNN). In specific, an improved segmentation algorithm is used to obtain the salient regions from the background as the proposals. To reduce the high false alarms from proposals, a lightweight CNN is used to classify these regions and make final predictions. Owning to the lack of current public infrared small target datasets, a new infrared dataset is proposed in this paper. The experimental results demonstrate that the proposed method has a good performance in detection rate and false alarm rate.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123493409","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}
Fine-grained visual classification (FGVC) aims to classify images belonging to the same basic category in a more detailed sub-category. It is a challenging research topic in the field of computer vision and pattern recognition in recent years. The existing FGVC method conduct the task by considering the part detection of the object in the image and its variants, which rarely pays attention to the difference in expression of many changes such as object size, posture, and perspective. As a result, these methods generally face two major difficulties: 1) How to effectively pay attention to the latent semantic region, and reduce the interference caused by many changes in pose and perspective; 2) How to extract rich feature information for non-rigid and weak structure objects. In order to solve these two problems, this paper proposes a deformable convolutional neural network with oriented response for FGVC. The proposed method can be divided into three main steps: firstly, the local region of latent semantic information is localized based on a lightweight CAM network; then, the deformable convolutional ResNet-50 network and the rotation-invariant coding oriented response network are designed, which input the original image and local region into the feature network to learn the discriminant features of rotation invariance; finally, the learned features are embed into a joint loss to optimize the entire network end-to-end. Experiments are carried out on three challenging FGVC datasets, including CUB-200-2011, FGVC_Aircraft and Aircraft_2 datasets. The results show that the accuracy of the proposed method on all datasets is better than the comparison method, which can effectively improve the accuracy of weakly supervised FGVC.
{"title":"A Deformable Convolutional Neural Network with Oriented Response for Fine-Grained Visual Classification","authors":"Shangxian Ruan, Jiating Yang, Jianbo Chen","doi":"10.1145/3457682.3457702","DOIUrl":"https://doi.org/10.1145/3457682.3457702","url":null,"abstract":"Fine-grained visual classification (FGVC) aims to classify images belonging to the same basic category in a more detailed sub-category. It is a challenging research topic in the field of computer vision and pattern recognition in recent years. The existing FGVC method conduct the task by considering the part detection of the object in the image and its variants, which rarely pays attention to the difference in expression of many changes such as object size, posture, and perspective. As a result, these methods generally face two major difficulties: 1) How to effectively pay attention to the latent semantic region, and reduce the interference caused by many changes in pose and perspective; 2) How to extract rich feature information for non-rigid and weak structure objects. In order to solve these two problems, this paper proposes a deformable convolutional neural network with oriented response for FGVC. The proposed method can be divided into three main steps: firstly, the local region of latent semantic information is localized based on a lightweight CAM network; then, the deformable convolutional ResNet-50 network and the rotation-invariant coding oriented response network are designed, which input the original image and local region into the feature network to learn the discriminant features of rotation invariance; finally, the learned features are embed into a joint loss to optimize the entire network end-to-end. Experiments are carried out on three challenging FGVC datasets, including CUB-200-2011, FGVC_Aircraft and Aircraft_2 datasets. The results show that the accuracy of the proposed method on all datasets is better than the comparison method, which can effectively improve the accuracy of weakly supervised FGVC.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124846733","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}
Detecting multi-scale pedestrians (especially small scale ones) is one of the most challenging problems in computer vision community. At present, most existing pedestrian detectors only adopt single-scale feature map in their backbone network for detection, which is not capable of fully taking advantages of multi-scale feature information, and thus resulting in unsatisfactory multi-scale detection performance. To address this issue, we propose in this paper a semantically enhanced multi-scale feature pyramid fusion method that can effectively extract and integrate multi-scale feature maps for multi-scale pedestrian detection. The proposed method consists of two main components: 1) Trapezoidal Path Augmented Module (TPAM) and 2) Multi-scale Feature Fusion Module (MFFM). TPAM aims at extracting higher-level semantic features by the additional higher-level feature layers, where the produced features are enhanced with supplementary higher-level semantic information, so that they can focus more accurately in the pedestrian area, leading to improved detection performance. MFFM aims at integrating multi-scale feature maps coming from TPAM to further take advantages of multi-scale feature information and reduce computational redundancy caused by multiple detection heads. By extensive experimental evaluations on the popular CityPersons and Caltech benchmarks, our proposed method achieves superior performances than previous state of the arts on multi-scale pedestrian detection.
{"title":"Semantically Enhanced Multi-scale Feature Pyramid Fusion for Pedestrian Detection","authors":"Jun Wang, Chao Zhu","doi":"10.1145/3457682.3457747","DOIUrl":"https://doi.org/10.1145/3457682.3457747","url":null,"abstract":"Detecting multi-scale pedestrians (especially small scale ones) is one of the most challenging problems in computer vision community. At present, most existing pedestrian detectors only adopt single-scale feature map in their backbone network for detection, which is not capable of fully taking advantages of multi-scale feature information, and thus resulting in unsatisfactory multi-scale detection performance. To address this issue, we propose in this paper a semantically enhanced multi-scale feature pyramid fusion method that can effectively extract and integrate multi-scale feature maps for multi-scale pedestrian detection. The proposed method consists of two main components: 1) Trapezoidal Path Augmented Module (TPAM) and 2) Multi-scale Feature Fusion Module (MFFM). TPAM aims at extracting higher-level semantic features by the additional higher-level feature layers, where the produced features are enhanced with supplementary higher-level semantic information, so that they can focus more accurately in the pedestrian area, leading to improved detection performance. MFFM aims at integrating multi-scale feature maps coming from TPAM to further take advantages of multi-scale feature information and reduce computational redundancy caused by multiple detection heads. By extensive experimental evaluations on the popular CityPersons and Caltech benchmarks, our proposed method achieves superior performances than previous state of the arts on multi-scale pedestrian detection.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127085800","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}
Ren Deng, Ye Liu, Liyan Luo, Dongjing Chen, Xijie Li
Feature selection is a machine learning technique that selects a representative subset of all features available in order to reduce the time and space needed to process high-dimensional data. Traditional feature selection methods include filter, wrapper, and embedded approaches. However, many conventional methods’ performances are not suitable in many contexts. This paper proposes a new unsupervised feature selection model based on pseudo label approximation. The new derived model incorporates a projection error, a sparsity regularization, and a manifold regularization term that preserves the manifold structure of the original data. Finally, implementation of the new model onto five distinct datasets validates the effectiveness of the proposed model.
{"title":"Unsupervised Feature Selection using Pseudo Label Approximation","authors":"Ren Deng, Ye Liu, Liyan Luo, Dongjing Chen, Xijie Li","doi":"10.1145/3457682.3457758","DOIUrl":"https://doi.org/10.1145/3457682.3457758","url":null,"abstract":"Feature selection is a machine learning technique that selects a representative subset of all features available in order to reduce the time and space needed to process high-dimensional data. Traditional feature selection methods include filter, wrapper, and embedded approaches. However, many conventional methods’ performances are not suitable in many contexts. This paper proposes a new unsupervised feature selection model based on pseudo label approximation. The new derived model incorporates a projection error, a sparsity regularization, and a manifold regularization term that preserves the manifold structure of the original data. Finally, implementation of the new model onto five distinct datasets validates the effectiveness of the proposed model.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125791390","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}
Hao-Tian Ren, Hongbo Zhang, Qinghongya Shi, Qing Lei, Jixiang Du
In this work, we discuss the feature fusion approach of multisurface skeleton projection images for action recognition. Multisurface skeleton projection images are generated from human skeleton joint motion trajectories on three surfaces: horizontal-vertical, horizontal- and vertical-depth surfaces. The vision features of these skeleton projection images contain complementary action information on different surfaces and are generally combined by early fusion or late fusion. To learn and fuse the discriminative features of each surface effectively, this paper proposes a new feature fusion method called the discriminative feature fusion network, which uses a two-task framework to recognize action and surface categories simultaneously. In the proposed network, the features of three skeleton projection images are computed by the same convolutional network. To retain the complementary action feature of these skeleton projection images, the surface classification loss is defined and added into the action classification loss to train the feature learning network. The experimental results show that the performance of the proposed feature fusion method is better than traditional early and late fusion. Compared with skeleton visualization image-based action recognition methods, the proposed method achieves state-of-art performance on the NTU RGB+D action dataset.
{"title":"DiffNet: Discriminative Feature Fusion Network of Multisurface Skeleton Project Images for Action Recognition","authors":"Hao-Tian Ren, Hongbo Zhang, Qinghongya Shi, Qing Lei, Jixiang Du","doi":"10.1145/3457682.3457749","DOIUrl":"https://doi.org/10.1145/3457682.3457749","url":null,"abstract":"In this work, we discuss the feature fusion approach of multisurface skeleton projection images for action recognition. Multisurface skeleton projection images are generated from human skeleton joint motion trajectories on three surfaces: horizontal-vertical, horizontal- and vertical-depth surfaces. The vision features of these skeleton projection images contain complementary action information on different surfaces and are generally combined by early fusion or late fusion. To learn and fuse the discriminative features of each surface effectively, this paper proposes a new feature fusion method called the discriminative feature fusion network, which uses a two-task framework to recognize action and surface categories simultaneously. In the proposed network, the features of three skeleton projection images are computed by the same convolutional network. To retain the complementary action feature of these skeleton projection images, the surface classification loss is defined and added into the action classification loss to train the feature learning network. The experimental results show that the performance of the proposed feature fusion method is better than traditional early and late fusion. Compared with skeleton visualization image-based action recognition methods, the proposed method achieves state-of-art performance on the NTU RGB+D action dataset.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126751389","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}
Yuze Qin, Guangzhong Sun, Jianfeng Li, Tianyu Hu, Yu He
With the development of computer science and technology, programming has become one of the college students’ essential abilities. The increasing number of students brings a big challenge to evaluating students’ programs. For saving human resources in checking code, most works attempt to design software to judge code automatically. However, these works focus on the best way to extract the semantic and syntax features of the correct programs, ignoring that judging for wrong programs is equally important to students. We design a grading model named SCG_FBS (Students’ Code Grading model, based on semantic Features analysis and Black-box testing, with a Select function) to extract semantic features of the code and evaluate codes based on the semantic features and black-box testing. We standardize the source code and translate it into a vector sequence by a pre-trained instruction embedding. Then we extract semantic features by a neural network with the attention method and concatenate semantic features with black-box testing results as the dependence for grading. Furthermore, we propose a select function to pick up significant sentences in each code, which can reduce the length of the input sequence and accelerate training. We gather two data sets from the OJ (Online Judge) platform, which is widely used in colleges to test students’ programs as a black-box. Our SCG_FBS model gets 87.92% accuracy on one data set and gets 84.28% accuracy on another. Meanwhile, our SCG_FBS model reduces 53.7% training time compared with baseline, significantly improving efficiency.
{"title":"SCG_FBS: A Code Grading Model for Students’ Program in Programming Education","authors":"Yuze Qin, Guangzhong Sun, Jianfeng Li, Tianyu Hu, Yu He","doi":"10.1145/3457682.3457714","DOIUrl":"https://doi.org/10.1145/3457682.3457714","url":null,"abstract":"With the development of computer science and technology, programming has become one of the college students’ essential abilities. The increasing number of students brings a big challenge to evaluating students’ programs. For saving human resources in checking code, most works attempt to design software to judge code automatically. However, these works focus on the best way to extract the semantic and syntax features of the correct programs, ignoring that judging for wrong programs is equally important to students. We design a grading model named SCG_FBS (Students’ Code Grading model, based on semantic Features analysis and Black-box testing, with a Select function) to extract semantic features of the code and evaluate codes based on the semantic features and black-box testing. We standardize the source code and translate it into a vector sequence by a pre-trained instruction embedding. Then we extract semantic features by a neural network with the attention method and concatenate semantic features with black-box testing results as the dependence for grading. Furthermore, we propose a select function to pick up significant sentences in each code, which can reduce the length of the input sequence and accelerate training. We gather two data sets from the OJ (Online Judge) platform, which is widely used in colleges to test students’ programs as a black-box. Our SCG_FBS model gets 87.92% accuracy on one data set and gets 84.28% accuracy on another. Meanwhile, our SCG_FBS model reduces 53.7% training time compared with baseline, significantly improving efficiency.","PeriodicalId":142045,"journal":{"name":"2021 13th International Conference on Machine Learning and Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129971872","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：