top of page

LATEST PROJECTS

 
 
Cost-Efficient Neurological Disorder detection ​

We proposed a neurological disorder detection scheme by considering the feature extraction cost. We encourage the classifiers to use cheap features instead of expensive ones. By doing that, we seek to reduce power consumption on a chip.

 
 
Model Compression on Oblique Trees

We proposed a model compression technique to generate sparse connected oblique trees. The model is compatible with cost-efficient inference.

Migraine Classification using Somatosensory Evoked Potentials

We proposed a system based on somatosensory evoked potentials (SSEP) for migraine state classification. Using gradient boosted decision trees, we achieved an accuracy of 89.7% by using a single-channel noninvasive SSEP.

bottom of page