TY - GEN
T1 - Lightweight and Interpretable Detection of Affective Engagement for Online Learners
AU - Boulanger, David
AU - Dewan, M. Ali Akber
AU - Kumar, Vivekanandan S.
AU - Lin, Fuhua
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - Offering online learners with a personalized, ethical, and privacy-protecting pedagogical companion that can detect their affective engagement and that can run on the devices they use to learn (edge devices) requires not only generating lightweight models but also interpretable, trustworthy, and generalizable ones. SqueezeNet is a lightweight deep neural network architecture, which is more popular for the edge devices. This paper showcases how SqueezeNet appears not to provide as meaningful explanations of its predictions of academic emotions as those of a lightweight traditional CNN model. More specifically, the winning CNN model, consisting of 473,317 parameters, delivers a predictive accuracy of 99.4% on the testing set, while maintaining an identical level of descriptive accuracy. In contrast, the best SqueezeNet model comprises 141,301 parameters (3x less) and has a predictive accuracy of 93.7% and a dropping descriptive accuracy of 84.0%. In brief, the SqueezeNet model does not maintain a similarly high descriptive accuracy and does not effectively identify facial features corresponding to the confusion, distraction, enjoyment, neutrality, and fatigue emotions.
AB - Offering online learners with a personalized, ethical, and privacy-protecting pedagogical companion that can detect their affective engagement and that can run on the devices they use to learn (edge devices) requires not only generating lightweight models but also interpretable, trustworthy, and generalizable ones. SqueezeNet is a lightweight deep neural network architecture, which is more popular for the edge devices. This paper showcases how SqueezeNet appears not to provide as meaningful explanations of its predictions of academic emotions as those of a lightweight traditional CNN model. More specifically, the winning CNN model, consisting of 473,317 parameters, delivers a predictive accuracy of 99.4% on the testing set, while maintaining an identical level of descriptive accuracy. In contrast, the best SqueezeNet model comprises 141,301 parameters (3x less) and has a predictive accuracy of 93.7% and a dropping descriptive accuracy of 84.0%. In brief, the SqueezeNet model does not maintain a similarly high descriptive accuracy and does not effectively identify facial features corresponding to the confusion, distraction, enjoyment, neutrality, and fatigue emotions.
KW - affective computing
KW - deep learning
KW - edge devices
KW - explainable artificial intelligence
KW - facial expressions
KW - online learning
KW - student engagement
UR - http://www.scopus.com/inward/record.url?scp=85127564393&partnerID=8YFLogxK
U2 - 10.1109/DASC-PICom-CBDCom-CyberSciTech52372.2021.00040
DO - 10.1109/DASC-PICom-CBDCom-CyberSciTech52372.2021.00040
M3 - Published Conference contribution
AN - SCOPUS:85127564393
T3 - Proceedings - 2021 IEEE International Conference on Dependable, Autonomic and Secure Computing, International Conference on Pervasive Intelligence and Computing, International Conference on Cloud and Big Data Computing and International Conference on Cyber Science and Technology Congress, DASC/PiCom/CBDCom/CyberSciTech 2021
SP - 176
EP - 184
BT - Proceedings - 2021 IEEE International Conference on Dependable, Autonomic and Secure Computing, International Conference on Pervasive Intelligence and Computing, International Conference on Cloud and Big Data Computing and International Conference on Cyber Science and Technology Congress, DASC/PiCom/CBDCom/CyberSciTech 2021
T2 - 19th IEEE International Conference on Dependable, Autonomic and Secure Computing, 19th IEEE International Conference on Pervasive Intelligence and Computing, 7th IEEE International Conference on Cloud and Big Data Computing and 2021 International Conference on Cyber Science and Technology Congress, DASC/PiCom/CBDCom/CyberSciTech 2021
Y2 - 25 October 2021 through 28 October 2021
ER -