Luis Costero

TL;DR: MAMUT, a multi-agent machine learning approach to tackle challenges of real-time video transcoding, consistently attains up to 8x improvement in terms of FPS violations, Quality of Service, and power reduction, as well as faster and more accurate adaptation both to the video contents and available resources.

TL;DR: This paper takes a different path that addresses the complexity of the problem at the library level, via a few asymmetry-aware fundamental kernels, hiding the architecture heterogeneity from the task scheduler, and shows that this elegant solution delivers much higher performance than a naive approach based on an asymmetric-oblivious scheduler.

TL;DR: A family of policies integrated within a runtime task scheduler (Nanox) pursue the goal of improving the energy efficiency of task-parallel executions with no intervention from the programmer by modifying the core operating frequency via DVFS mechanisms or by enabling/disabling the mapping of tasks to specific cores at selected execution points.

TL;DR: A multi-agent Reinforcement Learning framework to jointly adjust application- and system-level parameters at runtime to satisfy the QoS of multi-user HEVC streaming in power-constrained servers and it is shown that the power-capping techniques formulated outperform the hardware-based power capping with respect to quality.

TL;DR: This paper proposes strategies to build and serve this type of knowledge to concurrent applications by leveraging Reinforcement Learning techniques and results reveal an excellent adaptation of resource and knob management to heterogeneous QoS requests, and increases in the amount of concurrently served users.