In this repository, I have uploaded the projects I built while doing the Deep Learning Nanodegree from Udacity. I am enrolled in the Nanodegree after passing through phase 1 of a scholarship challenge course given by Udacity and Bertlesmann. This is also a phase 2 scholarship offered by them for selected performers of the first round of the scholarship.
In the first Nanodegree project I had to predict the number of bikes that would be needed from a bike-sharing company based on historical data and relevant features such as the humidity, wind, and day of the week. I implemented the forward, back-propagation, and update of weights in Numpy for the training of a fully-connected neural network. I also picked the hyperparameters that increased the accuracy the most. I achieved a validation loss of 0.168 based on the mean squared error of the last 60 days of the dataset.
In this project, I built a classifier based on Convolutional Neural Networks (CNNs) that is able to predict different breeds of dogs.
This project makes use of a Recurrent Neural Network (RNN) to create artificial scripts for a TV show after being trained on a script from the series The Office.
In this project, Generative Adversarial Networks (GANs) are used to create artificial faces after being trained on a dataset containing the faces of different celebrities.
In this project, I created an LSTM for sentiment analysis of movie reviews. The model was trained on AWS Sagemaker with the use of a dataset which contains reviews of movies from the website IMDB. Then, the trained model is deployed on a web app with the use of AWS services, such as Sagemaker, Lambda, and API Gateway.