Students guide to getting started with AI and Machine Learning

The reason I’m writing this post is I’ve started with my A.I Nanodegree course with Udacity and would love to share what I’ve learned so Far. This is the first of many posts to come that I will write as I progress further into this field.

There is a lot of heat around the words Machine Learning & A.I. but very few understand that Machine Learning is actually a subset of A.I. which deals more with feeding huge amount of data to the machine so that it learns from it and provide us with calculated outputs. A.I. is actually a very broad concept and mostly revolves around performing tasks in a manner that could be considered as more human like. But the current implementations of A.I. can only perform one task or a collection of similar tasks and are way behind for them to actually be considered human like. So, I started researching and here is my progress so far. Someone suggested me that firstly I should start with Three things and then start plotting the path for my career:

• Follow People who have done substantial work in the field.
• Follow top companies and what they use.
• Recent Research Papers or Breakthrough in the field.

So, Lets start with the First one, People. I personally found Sebastian Thrun, Peter Norvig & Yann LeCun. To describe them a bit:

Sebastian Thrun is the Chairman of Udacity and founded the Google X Project. He has done great work in the Field of Robotics and A.I. His course of A.I was one of the first MOOC courses which I took.

Peter Norvig is one of the world class A.I. Researchers and is the Director of Research at Google. He wrote THE Book on A.I. i.e. Artificial Intelligence: A Modern Approach.

Yann LeCun is the Director of AI Research at Facebook and Professor at NYU. He is a founding father of Convolutional Nets and one of the Leading Researchers on Computer Vision.

One more person I’d like to mention is Siraj Rawal. He is a Machine Learning and A.I. enthusiast with a Youtube Channel and is famous for covering the basics of a lot of things in a short time span. Some of you might find his videos too fast paced but for me it gives a push to newer concepts and ideas which I can research further.

P.S. Siraj has a Deep Learning Foundations Program with Udacity that covers topics like Neural Networks, Chatbots & Sentiment Analysis

Following these people and others has helped me a lot to understand what the field is growing into and what future it might hold.

Now let’s start with the Second one, Researching top companies and what they use. Big companies like Google, Facebook, IBM & Amazon either use and fund the top Libraries for A.I. Research and Machine Learning or Develop their own to solve bigger set of problems in their products or in the outside world. The Concept of using Libraries is very simple as it is better to solve the actual set of problems rather then developing a platform or methods which are used for generic purposes. Libraries like Tensorflow, Scikit Learn & Numpy are some of the examples. These Libraries can be used to train models and perform computations for almost anything. For example if a project needs to detect what type of flower is in a given picture, we don’t need to perform complex computations on the image. Rather we can use Prebuilt Image Classifier in Tensorlfow which can classify images based on various metrics of similarity and differences, train it with a given set of images and it will do the rest for us. More things to look for are Google Deepmind, OpenAI, IBM Watson & Facebook Artificial Intelligence Researchers (FAIR). I’ll dive deeper into these in my next posts.

Now moving on to the Third. In recent years A.I. has made great breakthroughs and has given us the glimpse of what we can achieve. The defeat of Lee Sedol by Alpha Go, Gary Kasparov by Deep Blue & the win at Jeopardy! by IBM Watson are one of the biggest breakthroughs that A.I. has made in the past years and showed how A.I. & Machine Learning can make computers think better than Humans at one of the most complex scenarios. Although these A.I. are far from being called Human like as they can still only excel at a single task. The current major fields what I found having the largest applications are Computer Vision & Natural Language Processing. There’s a lot of code available on GitHub and online to research further.

What I’ve learned so far in the Nanodegree?

Let’s start with what you need to dive. You need a basic knowledge of programming & preferably Python. A Knowledge of Statistics & Linear Algebra is a plus. If you’re a beginner I’d list the courses I found relevant and helpful:

• Intro to CS Course by Udacity or Harvard Online Learning
• Programming Foundations with Python at Udacity
• Linear Algebra by MIT
• Probability & Statistics Course by MIT
• Intro to A.I. course at Udacity
• Machine Learning Recipes at Google Developer YouTube Channel
• A.I. A Modern Approach by Peter Norvig
• Kaggle is also a good place to start playing with some Data Analytics

There is a great Repository on GitHub which can be found Here. It covers a whole path on how to become a Machine Learning Engineer and contained a lot of good material that has helped me study.

As I’ve mentioned I’m taking the A.I. Nanodegree course by Udacity, which is a 6 month long course split into two 3-month long terms. So far in the course we’ve made a A.I. Algorithm that solves Harder Sudoku Puzzles using Constraint Propagation & Depth First Search, the code to the project can be found Here. We also implemented Naked Twinks Strategy and wrote the code so that it solves the Diagonal Sudoku Problem. If you want to study more about the Sudoku Project follow Peter Norvig’s post Here. In the Second Project we’re making a Game Playing Agent that uses Min-Max Method and Alpha Beta Pruning Method for searching. The full course-ware for Term 1 can be found Here. I will be posting all my progress and what I learn though my journey so hang on tight.