Launch Into Machine Learning
This is the first time I decided to write something or more specifically express something in the context with the tech world. Through this, I would like to express my journey in the field of Machine Learning. Explicitly it will help enthusiasts to get a pathway.
For the starters, what excites one to get on board in the ship of machine learning is ample growth in opportunities, exponential development in the field over time, and also money(lots you receive :), hard work needed). What excited me is seeing the fascinating outcome where everything ranging from google keyboard to Amazon Alexa knows what we want to type next or which type of music we usually want to listen to. There’s an algorithm(s) behind making machines responding in such a smarter way. The backbone of these technologies is Artificial Intelligence. Oh, here we encountered a new term Artificial Intelligence(AI). The below image will give you a better understanding of domains.
I hope this image will help you clear your uncertainty in choosing the domain. Here, specifically, we’ll talk about the Machine Learning(ML) subset of AI. In this article, I’m not explaining machine learning, it’s more about how to learn it, which track to go along with.
Let's start with essentials.
Prerequisites Required
Basics of Python programming, a basic understanding of calculus, and concepts of Linear Algebra and Statistics. Many students flow in the hype of machine learning don’t follow the path. It’s highly advised by experience(I know a lot of freshers do this) not to do the mistake shown in the pic on left. One thing I want to elucidate that ML is more about mathematics than programming. Programming is assuredly convenient when you go for projects but mathematics is foremost and predominantly required for a strong base. The third step called algorithms is the foundation of machine learning models. To find hidden insights without needing explicit programs, machine learning uses algorithms that learn from previous data to help produce reliable and repeatable decisions. It’s critical to know the iterative part of machine learning, as models can have their very own psyche when presented to new, new data. Machine learning has “revolutionized” the universe of testing, and now, its algorithms can apply complicated mathematical calculations rapidly to enormous arrangements of data effectively and rapidly, on ordinary bases.
From Where to Begin
- For the starters, you have to outset from the Stanford University Machine Learning course by Andrew Ng. This course covers majorly used algorithms in ML with their proper mathematical interpretation, explained in a very intuitive aspect. This course is available on YouTube and Coursera. I can’t stop myself to say a few words about Andrew Ng. He’s a great competent instructor, and inspiration for all aspirants when it comes to Artificial Intelligence. Below is the frequently said line of him everyone admires a lot :).
In this course, MATLAB is used which nowadays is I consider outmoded for machine learning projects. So, for Python implementation of algorithms and understanding of libraries used you need to learn them separately.
- For learning basic Python there are many platforms and resources accessible, I’m not emphasizing that here. You can learn from anywhere. After basic python, you have to start learning libraries we’ll be using for the implementation of algorithms for practical purposes. Alongside the above(first point) course, you should start learning libraries. The course for this purpose I recommend is Python for Data Science and Machine Learning Bootcamp. This course gives insights on how to use algorithms in making ML models using python libraries. This course is available on Udemy and the instructor is Jose Marcial Portilla, proficient and skilled in explaining the approach of the subject.
- After achieving the milestone above, start working on small projects involving real-world datasets for conceding efficient implementation of the concepts acquired. There are many uncomplicated datasets available for novices to practice and employ ML methods. When it comes to operating with datasets, for getting data ready to serve the algorithm we need to prepare the data or more precisely(technical term) preprocess the data. Just like garnishing the food before serving increases the taste of food similarly performing data preprocessing, feature engineering, and visualization helps in analyzing data and increasing the accuracy.
- Datacamp is also an astounding platform for getting insights into machine learning. There are many subcourses available there to make the machine learning model perfect. Machine Learning Scientist is a course that covers all possible foreknowledge. You can get Datacamp premium subscription free for 3 months using the GitHub student developer pack.
- For the folks who prefer reading over watching lectures Medium is a classic spot to explore the charm of ML by reading. Towards Data Science and Analytics Vidhya are my exclusive preferences. You will discover the explanations by highly experienced individuals on these platforms.
Dive into Hands-on
An essential part of learning is applying principles in realistic situations. Kaggle is a platform for data science competitions where you get all the resources for doing hands-on. Kaggle kernels(notebooks) can be used to train models, the coolest part is we get free GPUs and TPUs used in training Deep Learning models. I started using Kaggle after I had significant experience(around a month) in the field. When you start working on first real-world data you will face situations like how to begin, how to visualize, and analyze data, that’s where the major learning stage starts. In the beginning, I faced the issue’s like I know the libraries but which one to use, what should be the structure or composition of the training model. At this time the best answer is to google it read about the techniques how you can make it work. To be honest, when I was working on my first dataset, I took help from other available kernels. Initially, you will see that different people are using different libraries, there are a lot of options available you need to choose among them and that’s where you start picking up the flow. Whenever I saw a new library or code or error I’m not able to resolve I just copy-pasted it and google, documentations are available, StackOverflow is another buddy ready to help you here.
By experience, I know there may be a situation where you are stuck or not getting any clue to walk ahead, or you are confused about what to start next, take advice from an endured person or you can ask me. As Elon Musk said:
“Persistence is very important. You should not give up unless you are forced to give up.”
Where to find me
Feel free to contact me. Always available to help if you are stuck somewhere in ML drive. I’m active on Linkedin. You can check my GitHub for projects related to ML, feel free to send a pull request, and contribute to projects. Here’s my Kaggle profile link.
