Resource Constrained Reasoning
Overview:
Everyday Sensing and Perception (ESP) is a megabet project in Intel
Research to drive research breakthroughs in sensing and inference
which enable a new class of context inference system that are "90%
accurate for over 90% of your day". To make such a system be applicable
to a wide range of applications and have big impact on our daily life,
the inference system should involve mobile (sensing) devices for data
processing, learning and action-taking, as they are becoming ubiquitous.
To support inference, reasoning and environmental learning on mobile
devices, we propose the research agenda in resource-constrained
reasoning, whcih focuses on developing efficient algorithms for signal
processing, feature extraction, object recognition, and other machine
learning based reasoning/inference/decision-making on mobile devices.
Because only limited computation power and communication bandwidth are
available on the mobile devices, the algorithms have to learn and
infer useful and high-level ideas with minimal computation and
communication cost.
In this project, we aim to develop algorithms which are efficient and
optimal under target accuracy requirement and resource constraints,
and study the fundamental trade-off between the degree of data
approximation and the achieved reasoning accuracy.
A set but not all research ideas in the resource-constrained reasoning
are as follows:
- Machine learning on compressively sensed data/compressed machine learning
- Efficient algorithms for object recognition in vision data
- Distributed detection and signal processing
- Machine learning, statistical inference and decision-making under
computation and communication constraints
Publications:
Talks: