Udacity Deep Learning Nanodegree Foundation Program

Synopsis

This is the repository of all the projects I will carry out during Udacity Deep Learning Nanodegree Foundation Program

Table of contents

Project 1 -- your first neural network

Using only numpy, we build a neural network from scratch to carry out predictions on daily bike rental ridership.

Project 2 -- images classification

Using tensorflow, we build a convolutional neural network to classify images from the CIFAR-10 dataset. The dataset consists of airplanes, dogs, cats, and other objects.

My network is as follows:

• 3 convolutionnal and max pool layers
• 1 flatten layer
• 3 fully connected layers with 0.5 dropout
• 1 output layer

My hyperparameters:

• Epoch: 20
• Batch size = 128

Project 3 -- TV Script Generation

In this project, we'll generate our own Simpsons TV scripts using RNNs. We'll be using part of the Simpsons dataset of scripts from 27 seasons. The Neural Network we'll build will generate a new TV script for a scene at Moe's Tavern.

My hyperparameters:

• num_epochs = 100
• batch_size = 64
• rnn_size = 512
• seq_length = 16
• learning_rate = 0.001
• show_every_n_batches = 128

Project 4 -- Language Translation

In this project, you’re going to take a peek into the realm of neural network machine translation. You’ll be training a sequence to sequence model on a dataset of English and French sentences that can translate new sentences from English to French.

My hyperparameters:

• epochs = 4
• batch_size = 256
• rnn_size = 128
• num_layers = 2
• encoding_embedding_size = 50
• decoding_embedding_size = 50
• learning_rate = 0.005
• keep_probability = 0.7

Project 5 -- Face Generation

In this project, you'll use generative adversarial networks to generate new images of faces.

My hyperparameters:

• MNIST

  • batch_size = 128
  • z_dim = 128
  • learning_rate = 0.0002
  • beta1 = 0.5
  • epochs = 2

• CelebA

  • batch_size = 128
  • z_dim = 100
  • learning_rate = 0.0001
  • beta1 = 0.5
  • epochs = 1

Prerequisites

You need:

• Jupyter
• Python 3.x
• The following Python library:
  • Numpy
  • Matplotlib
  • Tensorflow and all its dependencies

Running

To run the notebook, you need Jupyter installed with a Python 3.x kernel. Then just run

> jupyter-notebook-3.x notebook_name.ipynb

in the corresponding directory.