3 ms·
Why would machine learning be able to overcome a lack of statistical power?
by sillysaurus3 9y ago
Why would machine learning be able to overcome a lack of statistical power?
- cperciva 9y agoIt's not clear whether there's actually a lack of statistical power, or simply a lack of sufficient modelling of the noise.
- TeMPOraL 9y agoIt wouldn't, but maybe there is enough statistical power, but it's just tricky to process?
- harigov 9y agoDid you miss the part about sensitive sensors? A single sensor may not be enough to classify accurately but multiple sensors measuring different things can sometimes work. Such sensor fusion has worked incredibly well for IMU's and also in self-driving cars.
- lj3 9y agoThat's how the Glucotrack claims to work. From their site: GlucoTrack uses ultrasonic, electromagnetic and thermal technologies to non-invasively measure glucose levels in the blood. GlucoTrack is intended for use by Type 2 and pre-diabetics. GlucoTrack combines the following technologies: Ultrasound (speed of sound change within the tissue); Electromagnetic (conductivity of the tissue); Thermal (heat capacity of the tissue). It's not approved by the FDA, but it appears to be selling in Australia, New Zealand, Spain, China, Korea, Italy and some others.