In this work, we explore the utility of the three main types of neural networks: feed forward, convolutional, and recurrent. While using these networks, we develop a new way to model multiagent trajectory data, explore the use of multiple activation functions for neurons at each layer of a neural network, and create a new model that makes a dynamic number of predictions with sequence data.

Last modified

• 10/14/2019

Creator

• Mark Harmon

DOI

• https://doi.org/10.21985/n2-5eb6-3e41

Subject

• Engineering Sciences and Applied Mathematics

Keyword

• Stock Prediction
• Deep Learning
• Basketball Trajectory
• Activation Functions
• Machine Learning

Date created

• 2018-01-01

Resource type

• Dissertation

Rights statement

• In Copyright