My First Step Into Machine Learning
A guide to anyone who wants to learn machine learning from scratch, based on my experiences.
A few months back is when i started hearing a lot about machine learning and ai, I thought it was something that only M-Tech students and researchers dealt with. And for a long time that assumption had kept me away from approaching it. It was then that a lot of people around me started mentioning about the recent achievements in ai and i became curious about all that buzz that ai has been creating lately.
So , like any beginner , i dived into the topic with a google search.
First i explored the applications of ai. What ai can do in the real world and what real world problems it could solve.
And I was fascinated by a lot of things that i found. For example, how spotify uses machine learning to predict what songs users will like using the data set of the songs that the user repeatedly listens to.
And this article that tells how some guy from victoria recreated a large scale government project in just a few lines of code using neural networks.
and this ,
and many many more. Just open the Machine Learning tag on medium and you will be amazed with what you can find.
Reading all these articles only made me more and more interested in the topic of ai and i decided i’ll try a hand at learning it. So then began the search for online courses on ai. The first site I visited was
And it was actually a good site to start with . The site shows a detailed roadmap of where to start and what path to follow when taking your first step into machine learning. On searching more about the topic and also discussing with a few friends i found out about coursera and the machine learning course it was providing, and that too for free(not if you want to get certified). I started at once. As i completed the first week, i got hooked to the course.
This course is taught by Andrew Ng. Andrew Ng is VP & Chief Scientist of Baidu; Co-Chairman and Co-Founder of Coursera; and an Adjunct Professor at Stanford University.
If you ask me, if you want to start learning machine learning, then look no more. From what research i’v done searching for the right course, this is by far the best and the most highly rated. The course is 11 weeks long and each week is followed by an assignment in which you are made to implement all the algorithms that you learnt along the course of that week. This course uses Octave programming language and is very similar to python in terms of syntax.
As i completed week after week , my interest for the topic only increased. There is another site which teaches ml for beginners called fast.ai. But unlike the cousera course, fast.ai does not teach the underlying mechanisms of the different algorithms.
The great thing about the course is that, the exercises to be submitted at the end of each week is actually a real world application that has to be solved by us using the algorithm we learnt that week. That kept me motivated to complete the next week.
Week after Week some of the exercises that you’ll complete are :
# house price prediction in SF using real word data.
# hand written digits recognizer using a data set of images of digits from 0–9.
# classifying emails into spam and not spam.
# image compression by reducing the colors present in an image.
and more.
You will become familiar with the working of many of the most important algorithms in machine learning like :
# linear regression
# logistic regression
# neural networks
# k-means algorithm
# principle component analysis (dimensionality reduction)
I recently completed the course and believe me, i’v learnt more in the last three months than what college has taught me in the last 2 years. Next I plan on taking the deeplearning courses on coursera, also taught by Andrew Ng.
One mistake i made while taking the course was I went through the first few weeks really fast.Because of this, I dint get enough time to implement what i learnt on real world data. I later corrected this and it helped me a lot. I would suggest you to take an actual week to complete each week in the course, that way you will get the practice you require.
Also, after each week try implementing the algorithm you learnt on atleast one real world data set.
Great real world data sets can be found on :
and
This is my first blog post on medium. Feel free to criticize.
