Adolescents and their parents hesitate to acknowledge mental health issues until symptoms severely worsen, making timely treatment challenging. Moreover, infrequent psychiatric consultations often fail to adjust treatments to the dynamic nature of mental health states. To address these issues, our paper proposes an AI-based mental health assessment framework for adolescent mental health through non-invasively collected data from daily digital activities on their mobile devices, including tablets and smartphones. For this, we collect fifteen different types of passive sensor data across three primary categories of activities: studying, smartphone using, and metaverse gaming. Additionally, each adolescent completes self-survey reports on eight different disorders which are used as labels. Then, feature extraction is conducted based on this dataset, which yields 1,523 features that could function as potential digital biomarkers of mental health conditions in adolescents. Utilizing these features, our algorithm named CAMP: Customizable Automated Machine learning Process incorporates simulated annealing for feature selection. This approach enables the construction of AI models for mental health assessment that are finely tuned to domain specific strategies. Our experiments show that our proposed framework can significantly improve models' performance.

Targeting global EDF scheduling, this article proposes an optimal algorithm for parallelizing tasks with parallelization freedom. For this, we extend the interference-based sufficient schedulability analysis and derive monotonic increasing properties of both tolerance and interference for the schedulability. Leveraging those properties, we propose a one-way search–based optimal algorithm with polynomial time complexity. We present a formal proof of the optimality of the proposed algorithm. We first address the single-segment task model and then extend to the multi-segment task model. Our extensive experiments through both simulation and actual implementation show that our proposed approach can significantly improve the schedulability.

Real-time applications with high computational demand, e.g., autonomous driving, are emerging and their complex nature conforms to a DAG(directed acyclic graph) structure. We propose a conditionally optimal parallelization for real-time DAG tasks for global EDF, ensuring complete execution of all tasks within the deadline. To achieve this, we formalize a monotonic increasing property of both tolerance and interference to the parallelization option. Using such properties, we develop a unidirectional search algorithm that can assign parallelization options in polynomial time, which we formally prove the optimality. We observe significant improvement of schedulability through simulation experiment, and then in the following implementation experiment, we demonstrate that the algorithm is practically applicable for real-world use-cases.

Targeting global EDF scheduling, this paper proposes a conditionally optimal algorithm for parallelizing tasks with parallelization freedom. For this, we extend the interference-based sufficient schedulability analysis and derive monotonic increasing properties of both tolerance and interference for the schedulability. Leveraging those properties, we propose a one-way search based conditionally optimal algorithm with polynomial time complexity. Our extensive experiments through both simulation and actual implementation show that our proposed approach can significantly improve the schedulability up to 60 percent.

Recent parallel programming frameworks such as OpenCL and OpenMP allow us to enjoy the parallelization freedom for real-time tasks. The parallelization freedom creates the time versus density tradeoff problem in fluid scheduling, i.e., more parallelization reduces thread execution times but increases the density. By system-widely exercising this tradeoff, we propose optimal parameter tuning of real-time tasks aiming at maximizing the schedulability of multicore fluid scheduling. Our experimental study by both simulation and actual implementation shows that the proposed approach well balances the time and the density, and results in up to 80 percent improvement of the schedulability.