Researchers at the University of Nebraska-Lincoln (UNL) are on the cutting edge of machine-learning development, with UNL's Paul Quint investigating representation learning.
Quint says this process mainly concerns how a computer is thinking about data, and it extracts patterns by seeking to produce software that absorbs vast amounts of data and understands what the data means.
Quint's current effort combined a decision tree with a deep-learning model, and his team used a dataset of handwritten digits in an attempt to teach a machine to understand the relationship between their differences.
The decision tree helps address a common problem with artificial intelligence (AI), where it can be difficult to explain some of the conclusions an AI makes.
Such experiments are part of UNL's overarching push to explore the logical processes of AI deep learning, which Quint says is an important consideration in the quest to make machines interpretable by humans in order to ensure safety.
From Daily Nebraskan
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA