Speaker "Jeremiah Johnson" Details Back



Cheating at Deep Learning: Quick Prototyping using Keras


Getting satisfactory results from a deep learning model often requires experimentation and fine-tuning. This process can be time-consuming and slow, leading to less than optimal results. In this talk, we'll demonstrate how the high-level Keras library can be used to quickly build, train, and validate deep neural networks for a variety of different tasks, ranging from image analysis to natural language processing.


Machine learning researcher, Founder and Principal of Gorilla Data Science LLC, a data science consulting firm based in southern New Hampshire. Ph.D in mathematics (area of research computational abstract algebra). Expert predictive modeler. Research in deep learning and large-scale machine learning. I instruct courses in programming, computer science, mathematics, statistics, and analytics at the Bachelor and the Master level. Many of the courses I teach are part of the Analytics Bachelor of Science program at the University of New Hampshire - Manchester, an innovative, forward-thinking program that I developed and now direct. This program is one of only a handful of face-to-face undergraduate-level analytics programs in the nation. Fundamentally, I believe in scientific, data-driven decision-making, and I am passionate about encouraging the scientific approach to almost every endeavor.