“big data” machine“big data” machine learning for prediction and classification very large...
TRANSCRIPT
“Big data” machine learning for prediction and
classification
Daniel Acuna, Ph.D.
Rehabilitation Institute of Chicago & Northwestern University
“Elevator speech”
• Perfect storm for breakthroughs in machine learning
Very large unstructured datasets
Complex models Good prediction and classification
with little engineering
Increased computational power
Fast and accurate learning algorithms
• Typically done by a committee of experts based on education, awards, experience, publications, and funding.
Predicting scientific success
Quantitative measure of scientific “success”
• h-index: Measure of scientific “success”
Index of success using “big data”?
• Combining large datasets from heterogenous sources
Publications Funding Collaboration
Predicting scientific “success”
Acuna et al., Nature, 2012 “small data”
“big data”
Why “big data” works?
Simple model Complex model
Why “big data” works?
Simple model Complex model
Why “big data” works?
Why “big data” works?
Simple model Complex model
Why “big data” works?
Simple model Complex model
Why “big data” works?
“big data” “small data”
“Big data” machine learning for prediction and classification
Very large unstructured datasets
Complex models Good prediction and classification
with little engineering
Increased Computational power
Fast and accurate learning algorithms
Daniel Acuna, Ph.D.
Rehabilitation Institute of Chicago & Northwestern University