Alexa Researchers Improve AI Error Rate Up To 30% By Reducing Data Imbalance

In an attempt to mitigate bias from classifiers as a result of imbalanced training data, researchers at Amazon's Alexa division have developed a new technique to reduce error rates in some data-imbalanced systems.


This is a novel approach which can be seen as forming one part of the fight back against biased datasets, leading to performance improvements and more accurate results.


Read here.

#Bias #Amazon

©2019 by AITHICS