Deep Learning in Depth: Deep Learning versus Machine Learning
October 2018 • Podcast
In this podcast excerpt, Ritwik Gupta and Carson Sestili describe deep learning and how it differs from machine learning.
“As you compose more and more non-linear functions together, you can represent a much wider function space than you could with just one non-linear function. That is why deep learning is different from shallow learning. Shallow learning doesn’t compose multiple things together. Deep learning does.”
Software Engineering Institute
Although traditional machine learning methods are being successfully used to solve many problems in cybersecurity, their success often depends on choosing and extracting the right features from a dataset, which can be hard for complex data. In this podcast, Ritwik Gupta and Carson Sestili describe deep learning, a popular and quickly-growing subfield of machine learning that has had great success on problems about these datasets, and on many other problems where picking the right features for the job is hard or impossible. In this section of the podcast, the two researchers discuss how deep learning differs from machine learning.
About the Speaker
Ritwik Gupta is a machine learning researcher at the Carnegie Mellon University Software Engineering Institute’s Emerging Technology Center. His research focuses on the intersection of machine learning and health, with many forays into the areas of robotics, adversarial learning, and computational linguistics. He is passionate about educating people about machine learning and the many cool and unique ways it can be applied to unorthodox problem domains.
Carson Sestili is a machine learning research scientist in the CERT Data Science group, where he uses data science, statistics, and machine learning for research in cybersecurity and intelligence. His work at CERT has involved applying machine learning for problems in satellite image recognition, code security defect detection, and cyber-incident forecasting. He has also investigated machine learning models that identify novel and ambiguous data.