======================================================================
Date: 11/28 (Thu)
Time: 14:45-16:30
Place: 8-1-03
Speaker: Doreena Karmina Pulutan

Title:
Applying Deep Learning Techniques to Predict Torrential Rainfall in the Philippines

Abstract:
The mission of the ULAT project is to monitor torrential rainfall and typhoons in the Philippines using low-cost technology. We do this by deploying a dense network of cheap sensors to gather meteorological data all over the Philippines. Using the gathered data, I will use machine learning techniques to make short-term predictions on local torrential rainfall events. In particular, I will use convolutional long short term memory (convolutional LSTM) network, a deep learning technique that has shown high success in weather prediction applications. Convolutional LSTM is a kind of neural network that combines convolutional neural networks (CNNs) and LSTM, a special kind of recurrent neural network (RNN). In this talk, I will briefly discuss what machine learning is and its suitability for my intended application. I will then give an overview of how convolutional LSTM works and how it applies to the POTEKA data that we currently have.
======================================================================