We will have several shorter talks today instead of one big talk.
John Chan will give a talk on Zero Shot Learning, a classification problem using Deep Learning to achieve "Zero-Shot" Super-Resolution (SR). It is a first unsupervised CNN-based SR method. (following this paper https://arxiv.org/abs/1712.06087).
Orson Adams will give a talk on techniques for Named Entity Recognition:
HMMs and CRFs are well-known models used in sequence labeling tasks. He will briefly go over their mathematical properties and talk about why they are useful for Named Entity Recognition (NER). (https://en.wikipedia.org/wiki/Named-entity_recognition)
Combining LSTM with CRFs has been the state-of-the-art for the NER task for a while. We will talk about how they are combined, and we'll build a Bi-LSTM-CRF model using Keras (following this paper. https://arxiv.org/pdf/1603.01354.pdf).
Sean Reed will give a general survey on the use of Convolutional Neural Networks as applied to NLP tasks. He will present a few basic models in Keras as part of this survey.
About the Speakers
John Chan is a Data Scientist / Project Management consultant at KPMG Cognitive Technology Lab. His background is in FinTech and Data Analytics. He lives in Long Island.
Orson Adams is Data Scientist at Space Jam Data where he builds and maintains NLP pipelines. His background is in Applied Math and Statistics and he lives in Brooklyn. Space Jam Data has built a platform that crowdsources opinions (in the form of unstructured text) from millions of people about the retail options in their neighborhoods.
Sean Reed is a Data Scientist and Lead Instructor for Galvanize's Data Science Intensive Program, a program where dedicated students can start learning data science without giving up their day job. He has a Masters Degree in Economics from NYU and a Bachelors in Physics from Fordham University.