This blog introduces the various aspects of Machine Learning as a career. This gives guidance to those who have heard about Data Science but are confused about where to start. Many people are facing this situation due to the increasing trend in this path and I wanted to give a starter’s guidance of what I have experienced for the past 1 year.
What is Machine Learning?
Machine Learning is a field of study that gives computers the ability to learn without being explicitly programmed. What I mean by explicitly programmed is that we need not provide rules every time we use the computer. It now has the ability to learn!!
Let’s take an example:
All of us use Mail very often and there are categories such as Spam, Not Spam. Ever wondered How do Mail algorithms classify between these two?
- Just think about these questions. As Humans, how do we differentiate between the two? How long did we take to know that a mail is a spam or not? Initially, we learned by seeing each mail and then figure out that if the mail contains some specific words or advertisements, we know that its a spam. Similarly, a machine figures out a pattern when provided the data and then the next time it has to predict, it uses the previous prediction accuracy and the already available. This is what is meant by ‘not explicitly programmed’.
- Considering the same use case, let’s get some physical understanding of some terms. The Task of email spam classifier is to Classify Spam/Not Spam. Experience is the number of instances it has classified the mails before. A performance measure is the accuracy it has obtained in the previous experience. So if the performance on the task increases with experience, we can refer it to like learning and if it is the machine that is doing this, it is Machine Learning. So putting it all together, below is an official definition of what is Machine Learning.
“A Computer Program learns from Experience(E) with respect to some Task (T) and Performance measure(P), if its performance on Task(T) and Performance(P) improves with Experience(E).” — Tom M Mitchell.
Machine Learning Vs Artificial Intelligence
Artificial Intelligence (AI) and Machine Learning (ML) are the trending buzzwords and often many people get confused between the two. Many times they seem to be used interchangeably and this leads to confusion.
Artificial Intelligence is a broader concept of developing machines that are termed as ‘smart’. ML is a subfield of AI which enable machines to learn from past data or experience without being explicitly programmed. Artificial Intelligence — devices designed to be smart are classified into two fundamental groups — applied or general. Applied AI is most commonly used and these systems are commonly found in areas like Autonomous automobiles, etc.
Why do we need Machine Learning?
In the current generation where technology has no limits, huge amounts of data are being generated from various sectors of society. This data consists of images, audio, graphs, documents that are not being utilized and is Unstructured. When this data keeps increasing, it becomes harder to compute and make the best use of it. This is where Machine Learning comes into play where we can process huge amounts of data and make smart machines out of them.
Machine learning is a subset of Artificial Intelligence (AI) which aims at developing smarter machines by extracting this huge data. In our daily lives, we are not realizing that we use so many applications of ML wherever we go, be it office, homes, hospitals, transport sector, etc.
This image shows the rapid increase in data and prediction for the coming years. By 2025, there would be 175 ZB data that is generated, it is space equivalent to 4,334,613,046,576 Dvds!
Machine Learning Engineer = Countless Career opportunities
A career as a machine learning engineer is almost endless potential. It is ranked as the most-in-demand technical skill in today’s tech sector. According to Forbes, the global markets for services and products related to Machine Learning & Artificial Intelligence is expected to rise from $10.1 B in 2018 to $20.83 B by 2023. With Machine Learning’s job listing rising in areas such as Image Recognition, Natural Language Processing & deep learning, there is a wide number of opportunities regarding which specialty you are good at.
Is there a difference between Machine Learning & Conventional Programming?
Yes. In Conventional programming, we provide rules and data. The computer program provides back the answers or the consequence. Whereas, in machine learning, the computer program is provided with answers and the data. The output returned is the rules.
I will explain this with a brick game scenario. In the image, the code says that if the ball collides with a brick, brick vanishes, and the ball bounces back. These are the rules and the data we provide. The result is the game ie. answers.
Machine Learning rearranges this diagram as compared to traditional programming. Instead of us, as developers figuring out the rules such as when should the brick fall, when should the game end.., we could provide a bunch of examples on how the game should be, and then have the computer figure out the rules.
Confused from where to Start?
In my first year of college, as I was going through the online courses, I came across a course in Coursera which I found that many beginners have taken. I went through the syllabus and found that it is going to be a great course.
The course name is Machine Learning by Stanford University. This course is taught by Andrew Ng which makes it a perfect start for beginners. This learning path is an 11-week course where topics like Regression, Classification, Neural Networks, Support Vector Machines, etc are taught in-depth. The course has coding assignments using Matlab/Octave. Do not worry if you haven’t used either of these. You will get used it to it as you go through the course. The reason I feel that this course is great for Basics is that it does not directly focus on coding but rather gives strong concepts and then understand that concept by practice. Doing the assignments on your own will be a great start. This is the link to the course —
In the course, applications such as spam classifier, recommend movies based on the user such as Netflix gives a complete understanding to what extent ML applications are used.
I will suggest some more resources that will help increase skills. Deep Learning.ai has some learning material that is found in Coursera. The courses can be purchased using financial aid which is available in Coursera. DataCamp is another great site for learning fundamentals of ML. These courses give practice on applications including Image Classification, Prediction. Some commonly used programming languages are Python, R, Matlab, C++. I have found Google Colab to be great because of its interfacing and ease of use.
Kaggle is a great place to practice, compete and also download datasets. Kaggle has problems posed by many companies including Google,IBM, amazon which can lead you to some great opportunities.
Summary
For all those who have an interest in machine learning or curious about how Facebook, amazon gives you recommendations, your mobile camera detecting faces, and many more, this is the right time to make your curiosity and interest, a career and contribute to the ever-expanding technology.
Stay in tune for more!
If you have any queries regarding what next, or where to get more resources, contact me at pullarevuvit1145@gmail.com
