Recent advances in deep neural network architectures have enabled new techniques for time series classification and analysis. Using a simple data set as a running example, this talk discusses the merits of classical time series techniques, such as Wavelets, Fourier, PCA, ICA, and AR models. We then explore the merits of deep learning based time series techniques, such as CNN, RNN, and LSTM, with a particular emphasis on unsupervised methods, such as auto-encoders.
Dave Deriso is founder of SimpleHealth, which uses data science for preventative medicine. In academia, Dave was a lecturer at Stanford University's Department of Computational and Mathematical Engineering, was a Marie Cure fellow at Oxford Neuroscience, NSF undergraduate researcher at UCSD, and writer at Nature. Dave also led data science at Caviar, where he developed the intelligent routing system for order deliveries, before it sold to Square. He holds an MS in Computational and Mathematical Engineering / Data Science from Stanford.