Udacity ML projects

I decided to upload my Udacity MLND projects to this blog. I think some of my answers in notebooks are really useful and they perhaps give any insights to you.

Note: The report will be redirected to nbviewer because rendering with nbviewer is more beautiful than that of the github engine.

Note2: If you want to grasp the real datasets, feel free to download them from official respostories

Boston housing

Full report | Github repo

In this project, you will apply basic machine learning concepts on data collected for housing prices in the Boston, Massachusetts area to predict the selling price of a new home. You will first explore the data to obtain important features and descriptive statistics about the dataset. Next, you will properly split the data into testing and training subsets, and determine a suitable performance metric for this problem. You will then analyze performance graphs for a learning algorithm with varying parameters and training set sizes. This will enable you to pick the optimal model that best generalizes for unseen data. Finally, you will test this optimal model on a new sample and compare the predicted selling price to your statistics.

Finding donors

Full report | Github repo

In this project, you will apply supervised learning techniques and an analytical mind on data collected for the U.S. census to help CharityML (a fictitious charity organization) identify people most likely to donate to their cause. You will first explore the data to learn how the census data is recorded. Next, you will apply a series of transformations and preprocessing techniques to manipulate the data into a workable format. You will then evaluate several supervised learners of your choice on the data, and consider which is best suited for the solution. Afterwards, you will optimize the model you’ve selected and present it as your solution to CharityML. Finally, you will explore the chosen model and its predictions under the hood, to see just how well it’s performing when considering the data it’s given. predicted selling price to your statistics.

Customer segments

Full report | Github repo

In this project you will apply unsupervised learning techniques on product spending data collected for customers of a wholesale distributor in Lisbon, Portugal to identify customer segments hidden in the data. You will first explore the data by selecting a small subset to sample and determine if any product categories highly correlate with one another. Afterwards, you will preprocess the data by scaling each product category and then identifying (and removing) unwanted outliers. With the good, clean customer spending data, you will apply PCA transformations to the data and implement clustering algorithms to segment the transformed customer data. Finally, you will compare the segmentation found with an additional labeling and consider ways this information could assist the wholesale distributor with future service changes.

Teach a Quadcopter How to Fly!

Full report | Github repo

In this project, you will design an agent to fly a quadcopter, and then train it using a reinforcement learning algorithm of your choice - The DDPG actor-critic method was what I chose and it went very well !

Dog breed classifier

Full report | Github repo

Welcome to the Convolutional Neural Networks (CNN) project in the AI Nanodegree! In this project, you will learn how to build a pipeline that can be used within a web or mobile app to process real-world, user-supplied images. Given an image of a dog, your algorithm will identify an estimate of the canine’s breed. If supplied an image of a human, the code will identify the resembling dog breed.

sample-dog-output

Along with exploring state-of-the-art CNN models for classification, you will make important design decisions about the user experience for your app. Our goal is that by completing this lab, you understand the challenges involved in piecing together a series of models designed to perform various tasks in a data processing pipeline. Each model has its strengths and weaknesses, and engineering a real-world application often involves solving many problems without a perfect answer. Your imperfect solution will nonetheless create a fun user experience!