My experience with Udacity’s Machine Learning Nanodegree Program

-Divyansh Sharma

I graduated from Udacity Machine Learning Nanodegree in July,2017 and I will be sharing how it has helped me learn,grow and apply my gained nanodegree skills to real world problems.

What is Nanodegree anyway?

A regular degree from a university takes around 4 years to complete. A regular degree also has a few core courses, some electives, and some open-ended projects. Regular degrees provide certification to signal that the student is ready to work in a field.

A Nanodegree program is like a regular degree in the sense that it also has some core parts and electives. But the timeline for a Nanodegree program is much smaller, around 6–12 months. (hence ‘nano’). Udacity partners with different companies like Kaggle, Facebook, and Google for creating courses. Each Nanodegree program finishes with a capstone project. Students apply their knowledge gained in the coursework to the capstone in novel ways.

My Background

I am a software developer having 2 years of experience in software development.My interest in machine learning piqued in final year of my engineering degree in which i was making a project using machine learning techniques.Since then i am continuously adding machine learning skills to my portfolio.But until completing my nanodegree i had only theoretical knowledge of machine learning.Udacity’s MLND ensures that each term finishes with a project so as to give the student practical application of the gained skills.

Prerequisites

Udacity lists their prerequisites on the MLND homepage.

Program Structure

The Machine Learning Nanodegree program is now split up in two sub modules which was not there earlier.It facilitates a beginner to quickly pick up the track and an advanced to quickly choose the right course needed.

>Basics : Contains two technical project with minimal prerequisites.

>Advanced : Contains three technical projects including one capstone project.

Each project has video lectures and in-lecture quizzes for practice.Each project has its own rubric. Reviewers grade projects based on these rubrics. Students can submit as many times as they like. Project submission is either via Github repository or zipped files.

What is unique to Udacity is the peer review system in which the reviewers not only review your project submission but also give you additional friendly comments on strategy,machine learning trends and resources to read more about a particular topic of the project.Each project has a timeline duration in which you have to submit the project as github link or zipped files.

General Projects

I’ll briefly cover my version of the syllabus which may be different from what is there now as Udacity keeps on reviewing and updating the project and course content to keep it industry ready.

Each project container boilerplate python code in jupyter notebooks with instructions step by step so that you can start off quickly and code only what is required.

P0 : Titanic Survival Exploration(optional)

This is an optional project using the Titanic dataset from Kaggle. It’s a classification problem where we have to predict the survival outcome of passengers. It’s supposed to get students familiar with the submission and review cycle.

P1: Predicting Boston Housing Prices

This is one of the mandatory projects. It uses supervised learning techniques to predict the price of houses in the Boston area. It’s a classic dataset from the UCI Machine Learning Repository . The goal is to choose the best performing model using supervised learning.

P2 : Building a Student Intervention System

This is a classification problem where the task is to identify students who might need early intervention before they fail to graduate.This project has been replaced by a newer version in current curriculum.

P3: Creating Customer Segments

Unsupervised learning techniques like PCA and clustering are explored here with the wholesale customers dataset from the UCI repository.

P4: Train a Smartcab to Drive

The lessons in this project focus on reinforcement learning. The goal is to take a smartcab from a starting point to an end point by giving it feedback through reinforcement learning.

P5 : Classify images

This project implements deep learning methods to CIFAR-10 dataset using Tensorflow.

Capstone

The capstone is the final technical project in the MLND. It’s divided into two parts: proposal and project implementation. Students are supposed to choose a technical problem from a domain they are interested in, write-up a proposal, solve the problem using machine learning techniques, and submit a final report for the project.

This capstone project is what unique to Udacity.The proposal and implementation parts lets you document your project and your approach process so that you actually do a research on a particular topic just like writing a research paper.Its thoroughly reviewed by Udacity reviewers with friendly comments on places to improve.

I worked on the Kaggle Invasive species monitoring competition using Keras for my capstone. It was a binary image classification problem.It took me a lot of research on which model to use which technique to use and document it in proper format as given by the capstone project rubric.But this helped get to know the research paper writing process and its implementation and later I published my own research paper in a national conference on machine learning.

Certification

Students who successfully complete all the program requirements earn and receive their Nanodegree credential.

Support

Each Nanodegree program has its own discussion forums. MLND also has a slack channel where many former graduates give advice. Also with new Udacity 1 to 1 mentorship be it a small doubt about course or be it if you are stuck in your projects you will always find someone to help.Also the project reviewers comments on your submitted projects gives you their experienced knowledge that you will not find easily.

Overall

My experience with this program has helped me to learn how to research, experiment, and finish an open-ended machine learning project on my own. It taught me how to start a project from scratch,document it and explain it to someone else.As said earlier the reviewers were incredibly helpful and gave leads to many great resources that you won’t find just searching.

Enroll Now into Machine Learning Nanodegree here

About the Author | Divyansh Sharma

Engineer by profession,Writer by heart and Musician by soul.