摘要:
Learning discriminative features for visually similar classes is crucial for fine-grained image recognition tasks. Bilinear pooling models use the outer product of embedding features to enhance the representation capability and achieve favorable classification performance. However, these models cause exceedingly high dimensionality of features which makes them impractical for large-scale applications and may result in overfitting. This article proposes a feature correlation residual method to mine the channel and spatial correlation of embedding features without increasing the dimensionality of features. For this purpose, each channel/location of the embedding features in the residual module is determined by its channel/spatial correlation to all other channels/locations. Then, the correlation residual features are used to complement the original ones. In addition to cross entropy loss, batch nuclear norm loss and triplet loss based on the extracted features are used as regularization to alleviate overfitting, enlarge inter-class variations and reduce intra-class variations. Experimental results show that our method achieves state-of-the-art performances on some popular datasets for fine-grained image recognition.
作者机构:
[Pang, Feng; Gu, Wei-Jiang] Information Center, Nanjing Forestry University, Nanjing;210037, China;[Lu, Kai-Li] National Engineering Center for E-Learning, Central China Normal University, Wuhan;430079, China;[Pang, Feng; Gu, Wei-Jiang] 210037, China
会议名称:
5th International Conference on Distance Education and Learning, ICDEL 2020
作者机构:
[Xiang, Wei] School of Educational Information Technology, Central China Normal University, Wuhan, Hubei, China;[Jing, Xiuji] School of Foreign Language, Central China Normal University, Wuhan, Hubei, China
会议名称:
3rd International Conference on Education Technology Management, ICETM 2020
作者机构:
[Yang, Jiumin; Xu, Ke; Pi, Zhongling] Cent China Normal Univ, Sch Educ Informat Technol, Wuhan 430079, Hubei, Peoples R China.;[Liu, Caixia] Cent China Normal Univ, Collaborat Innovat Ctr Informat Technol & Balance, Wuhan 430079, Hubei, Peoples R China.;[Yang, Jiumin] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan 430079, Hubei, Peoples R China.;[Yang, Jiumin] Cent China Normal Univ, 152 Luoyu Rd, Wuhan 430079, Hubei, Peoples R China.
通讯机构:
[Yang, Jiumin] C;Cent China Normal Univ, 152 Luoyu Rd, Wuhan 430079, Hubei, Peoples R China.
摘要:
The paper presents the research and comparative analysis of the bandwidth of low-power wireless IoT devices as wireless switches. Investigated and analyzed sensors the dependence of FTP multimedia data transmission speed on wireless Wi-Fi network on the temperature of the switch processor. To get temperature indicators sufficiently built into Python libraries to read temperature files. The paper focuses on the synchronicity of measurement results records for more accurate analysis. As a result, the dependence of the measured factors was calculated using the Pearson correlation formula. These measurement factors affect the autonomy and energy consumption, which is important for IoT devices, and therefore, among the devices tested, recommendations were made regarding their choice when used depending on the conditions.
摘要:
In recent days, the wireless communication technology has become an integral part of several types of communication devices as it allows users to communicate even from remote areas. Wi-Fi is one of the main types of wireless communication used by many electronic devices such as laptops, smart phones and Internet of Things (IoT) devices. Using Wi-Fi infrastructure to provide quality multimedia services and deploying IoT services requires new approaches to network resource management. The method of adaptive size formation of a Wi-Fi network cell in the corporate infrastructure is proposed. The idea of the method is to adapt the size of a Wi-Fi cell coverage depending on the QoS requirements and load localization under conditions of insufficient bandwidth resources for network users and IoT devices. This method is implemented by writing a script file on the Cisco 5508 Wi-Fi controller that makes informed decisions on changing the cell coverage size in a centralized manner. Using this method allowed to provide high-speed wireless Internet access for users, provide a good QoS for IoT services and increase Wi-Fi network availability in high network-load conditions by effectively managing network resources depending on the actual operation scenario and requirements of individual users and IoT devices.
期刊:
Advances in Intelligent Systems and Computing,2020年938:13-22 ISSN:2194-5357
通讯作者:
Radchenko, K.
作者机构:
[Hu Z.] School of Educational Information Technology, Central China Normal University, Wuhan, China;[Tereikovskyi I.; Radchenko K.] National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv, Ukraine;[Tereikovska L.; Tsiutsiura M.] Kyiv National University of Construction and Architecture, Kyiv, Ukraine
通讯机构:
[Radchenko, K.] N;National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”Ukraine
会议名称:
2nd International Conference on Computer Science, Engineering and Education Applications, ICCSEEA 2019
会议时间:
26 January 2019 through 27 January 2019
关键词:
Forecasting;Load;Mother wavelet;Wavelet model;Web server
摘要:
Metacognition is important in self-regulated learning and understanding its epistemic network can improve teaching and learning. We collected self-reported metacognition reflections on collaborative learning activities from 87 college students to analyze how students' metacognitive patterns differ by performance level and discipline type. We used an epistemic network analysis to identify these differences, and the results indicated that description of goals appeared most in self-reported reflections. There are variations in metacognitive patterns between different groups. High-score students had stronger connections around actions, while low-score students had stronger connections between metacognitive knowledge and context. The natural science students focused more on metacognitive knowledge and actions, while the humanities science students focused more on metacognitive experience and context. This implies that teachers should provide clear explanations about the collaborative learning goal, and a group strategy that takes both performance and discipline types into consideration could address the variation in metacognitive patterns.
摘要:
This study analysed the instructors' teaching presence of three courses conducted by an instructor to explore the effects of the instructors' online teaching presence on students' interactions and collaborative knowledge constructions. Content analysis, social network analysis, and lag sequential analysis were used to explore the mechanism of teaching presence on students' interactions and collaborative knowledge construction. Results demonstrate that the design and organization, as well as facilitating discourse, can facilitate students' interaction, reduce the number of peripheral students, and facilitate students' collaborative knowledge construction, especially in the knowledge sharing, discovery, discussion, and application, whereas direct instruction has positive effects on teachers' centrality and negative effects on knowledge negotiation and testing. The result can give the instructors some guidance on online teaching practices.
What is already known about this topic
What this paper adds
摘要:
This study investigates the role of a collaboration script, the Funnel Model, in supporting students’ computer-supported collaborative scientific argumentation, and how the students appropriated the collaboration script in scientific argumentation. In this exploratory case study, a class of 33 Secondary grade four students went through four phases of computer-supported collaborative argumentation activity scripted by the Funnel Model: individual ideation, intra-group synergy, inter-group critique and intra-group refinement. Multiple sources of data were collected including student-generated artefacts online at different phases of collaboration, and the post-intervention interviews with the students. The results show that the Funnel Model facilitated students’ computer-supported collaborative argumentation. The students’ levels of content mastery, motivation, classroom culture and time allocated for classroom participation affect students’ appropriation of the script for effective collaborative argumentation.