Artificial Intelligence Engineer Nanodegree: Deep Learning Applications

Recurrent Neural Networks

Time Series Prediction and Text Generation

Used Recurrent Neural Networks (RNN) to build RNNs that can generate sequences based on input data - with a focus on two applications: With the first I used real market data in order to predict future Apple stock prices using an RNN model. The second one is trained on Sir Arthur Conan Doyle's classic novel Sherlock Holmes and will generate wacky sentences based on it that may - or may not - become the next great Sherlock Holmes novel!

Code

• RNN_project.ipynb - Interactive Python Notebook.
• report.pdf - Report of Python Notebook.
• report.html - Report of Python Notebook.

Getting Started

To get this code on your machine you can fork the repo or open a terminal and run this command.

$ git clone https://github.com/JonathanKSullivan/text_generation.git
$ cd text_generation
$ jupyter notebook RNN_project.ipynb

Prerequisites

This project requires Python 3:

Notes:

1. It is highly recommended that you install the Anaconda distribution of Python and load the environment included below.

Installing

Mac OS X and Linux

1. Run git clone https://github.com/udacity/aind2-dl.git; cd aind2-dl
2. Run conda env create -f requirements/aind-dl-mac-linux.yml
3. Run source activate aind-dl
4. Run KERAS_BACKEND=tensorflow python -c "from keras import backend"

Windows

1. Run git clone https://github.com/udacity/aind2-dl.git; cd aind2-dl
2. Run conda env create -f requirements/aind-dl-windows.yml
3. Run activate aind-dl
4. Run set KERAS_BACKEND=tensorflow
5. Run python -c "from keras import backend"

Built With

• Anaconda - The data science platform used

Authors

• Udacity - Initial work - AIND-Isolation
• Jonathan Sulivan

Acknowledgments

• Hackbright Academy
• Udacity