A5044119432- Interconnection Networks and Systems
- Age of Information Optimization
- Parallel Computing and Optimization Techniques
- Advanced Neural Network Applications
- Privacy-Preserving Technologies in Data
- Green IT and Sustainability
- Distributed systems and fault tolerance
- IoT and Edge/Fog Computing
- Brain Tumor Detection and Classification
Johns Hopkins University
2021-2023
Programmable packet scheduling enables algorithms to be programmed into the data plane without changing hardware. Existing proposals either have no hardware implementations for switch ASICs or require multiple strict-priority queues.
The need to train DNN models on end-user devices (e.g., smartphones) is increasing with the improve data privacy and reduce communication overheads. Unlike datacenter servers powerful CPUs GPUs, modern smartphones consist of a diverse collection specialized cores following system-on-a-chip (SoC) architecture that together perform variety tasks. We observe training DNNs smartphone SoC without carefully considering its resource constraints can not only lead suboptimal performance but...
Training DNNs on a smartphone system-on-a-chip (SoC) without carefully considering its resource constraints leads to suboptimal training performance and significantly affects user experience. To this end, we present Flamingo, system for smartphones that optimizes DNN time energy under dynamic availability, by scaling parallelism exploiting compute heterogeneity in real-time. As AI becomes part of the mainstream experience, need train on-device crucial fine-tune predictive models while...