Search

Multi-task learning in Convolutional Neural Networks (CNNs) has led to remarkable success in a variety of applications of computer vision. Towards effective multi-task CNN architectures, recent studies automatically learn the optimal combinations of task-specific features at single network layers. However, they generally learn an unchanged operation of feature combination after training, regardless of the characteristic changes of task-specific features across different inputs. In this paper, we propose a novel Adaptive Feature Aggregation (AFA) layer for multi-task CNNs, in which a dynamic aggregation mechanism is designed to allow each task adaptively determines the degree to which the knowledge sharing or preserving between tasks is needed based on the characteristics of inputs. We introduce two types of aggregation modules to the AFA layer, which realize the adaptive feature aggregation by capturing the feature dependencies of different tasks along the channel and spatial axes, respectively. The AFA layer is a plug-and-play component with low parameter and computation overheads, and can be trained end-to-end along with backbone networks. For both pixel-level and image-level tasks, we empirically show that our approach substantially outperforms the previous state-of-the-art methods of multi-task CNNs. The code and models are available at https://github.com/zhenshen-mla/AFANet. [ABSTRACT FROM AUTHOR]

In higher education, the initial studying period of each course plays a crucial role for students, and seriously influences the subsequent learning activities. However, given the large size of a course’s students at universities, it has become impossible for teachers to keep track of the performance of individual students. In this circumstance, an academic early warning system is desirable, which automatically detects students with difficulties in learning (i.e., at-risk students) prior to a course starting. However, previous studies are not well suited to this purpose for two reasons: 1) they have mainly concentrated on e-learning platforms, e.g., massive open online courses (MOOCs), and relied on the data about students’ online activities, which is hardly accessed in traditional teaching scenarios; and 2) they have only made performance prediction when a course is in progress or even close to the end. In this paper, for traditional classroom-teaching scenarios, we investigate the task of pre-course student performance prediction, which refers to detecting at-risk students for each course before its commencement. To better represent a student sample and utilize the correlations among courses, we cast the problem as a multi-instance multi-label (MIML) problem. Besides, given the problem of data scarcity, we propose a novel multi-task learning method, i.e., MIML-Circle, to predict the performance of students from different specialties in a unified framework. Extensive experiments are conducted on five real-world datasets, and the results demonstrate the superiority of our approach over the state-of-the-art methods. [ABSTRACT FROM AUTHOR]