Plan to learn Machine Learning/Data Science in 2021? Note these assets from 2020.
“Before we start to dive deeper into the topic, let us first understand why do we need Machine Learning & ask yourself a question, do you really need to learn Machine Learning? If so, then why?”
(Note: These are all my opinion so, have them as a reference but, decide based on what suits you well)
A Caution: If it only because of the hype around the technology in a recent time of the pandemic, let me warn you that, no doubt it is a quite rewarding career, but it takes a considerable amount of time, the dedication & perseverance to stay longer in the field. The technology is evolving at a rate where what you might learn today may be obsolete or less likely be used in the practice in the future, remember the knowledge stays forever. So, if you are determined to follow religiously the career path & willing to update yourself by learning constantly, then I heartily welcome you & congratulate as you just accomplished a significant challenge, which is to set a mindset to follow this as a career path.
Suggestion: If you are an undergraduate, you have the most useful resource which life can provide you with, you guessed it right, “Time” so, utilize it to explore the domain & strengthen your fundamentals. If you are a working professional, not a wise decision to leave the job & learn ML unless you are planning for higher studies. Your experience will pay off over time. All you need is to learn “Prioritizing” & “Scheduling” tasks, not “Procrastinating.” With this, let’s begin with first understanding:
What is Machine Learning (ML)?
To put it in simple words, it is all about “M”, not “Magic” but, “Mathematics.” Don’t be afraid, but be prepared to get into the world of mathematics to really understand the technology. Let’s just keep it simple for the day & move on to the typical definition, which is:
Wikipedia: Machine learning (ML) is the study of computer algorithms that improve automatically through experience. It is seen as a subset of artificial intelligence (AI).
Without any further ado, let’s get into the part of choosing the right way to learn Machine Learning.
Remember, told you to be prepared for something. Here it comes & you guessed it right:
1. Mathematics & Statistics:
Machine learning is built on Mathematical prerequisite & creating an ML model is all about creating an algorithm that can learn from the data very well & generate satisfactory results on unseen data. (usually, we use algorithms already created but, if you can build a new one, that’s quite what researchers spend time on, in developing new methods using this knowledge.)
“We can always create a model which gives you good results but, how do you know if it’s the best or where to use it? Here comes the knowledge of mathematics which lets you think beyond typical solutions”
What to learn in Mathematics for Machine Learning?
Well, you need to learn the right concept & it is pretty much understandable (basic concepts, though) if you are good with your high school mathematics, if not, go back to those NCERT or any other textbook & start, no other way around to be here. The depth of the concepts truly depends upon the kind of role you’d be playing at any organization, but you need to start somewhere. “Told you to be prepared.”
Follow:
1. Linear Algebra
2. Calculus
3. Statistics
4. Probability
I won’t be going to the depth of each topic, rather provide you with the right resources (to the best of my knowledge) to learn from.
Where do I learn from?
Online Courses
1. Math track of Khan Academy: Linear Algebra, Probability & Statistics, Multivariable Calculus and Optimization.
2. Linear Algebra by MIT Open Courseware by Prof. Gilbert Strang- Considered to be one of the best lectures out there to learn from.
“Feel free to follow whichever suits your style.”
3. Statistics:
- Khan Academy videos on Statistics.
- Intro to statistics by Udacity.
- If you have a good amount of time for Probability fundamentals, random variables, probability distributions, hypothesis testing, etc. then I can recommend this detailed Georgia Tech’s course “ Statistical Methods”. An intuitive way to learn probability & other related concepts.
- Seeing theory: One of the best resources to understand the basics of probability intuitively. (Recommended)
Books to refer:
- Mathematics for Machine Learning by Marc Peter Deisenroth, A. Aldo Faisal, Cheng Soon Ong: To learn the concepts behind the technology. (MML book) (open in the browser if needed).
2. An Introduction to Statistical Learning: An Application with R: ISLR Whether you choose from Python/R, this book is to get the statistical knowledge.
3. Open Intro Statistics: Set the price to $0 to get it for free else, you can always contribute to the author as per your wish.
4. Refer the book for in-depth knowledge: Think Stats.
5. The Elements of Statistical Learning: Data Mining, Inference, and Prediction.
“There are many other available resources/books but, would like to keep things simple & not overwhelm with the content.”
2. Machine Learning, know where to start from?
To get started with the introduction to the world of Machine Learning, you can refer to these courses & books:
Online Courses
1. Coursera: Machine learning by Andrew Ng.
The course is free but no certification upon completion but, you can always apply for a Financial Aid (usually takes 2–3 weeks to approve) Note: Do not start the audit for free if applying for financial aid.
2. MITOpenLearningLibrary: Introduction to Machine Learning (6.036) — A 13-week program offered for free. No certificates, only for learning purpose.
3. Coursera: Practical Machine Learning by John Hopkins University.
4. Edx: Machine Learning from Columbia University.
5. Explore ai.google for more exciting content & to learn from ML experts at Google.
Books to Refer:
1. Deeplearningbook: Now, you may be wondering where this Deep Learning came from? Just refer the first 5 chapters for ML content with in-depth mathematics. Must read an end-end book, if planning for Machine Learning engineer roles (let’s talk about various roles in another article.)
2. Pattern Recognition and Machine Learning Book by Christopher Bishop.
3. Learn to Code
Wait! “did someone tell you that, you can become a Data Scientist/ Machine Learning Engineer without coding knowledge?” — A debatable topic however, it is good to have knowledge of coding as this is often what’s going to get you past the Interviews & strengthen your candidature!
- Learn about computer science fundamentals. You can directly jump to learn any of the programming languages but good to have a base understanding of the field/subject & you can visit CS50x to have a good start. For verified certification visit edx for some price but have an option available for financial aid (search for it online on how to get this).
2. Choose any of the preferred languages.
- a) Open Source: Python/R, etc. Usually suggest python, start with as it is user friendly & has good libraries available with strong community available for help whenever stuck. But R is more focused when the analysis is inclined towards statistics more. Decide as per the industry/target companies you would like to work with or have your own start-up, the choice is yours.
- b) Commercial: SAS, SPSS, etc.
3. Where to learn to program from? Well, here I can’t guarantee you any justice as there’s a plethora of information available in this regard. I Can suggest a few to you but, do not limit yourself to these options & at the same time not be overwhelmed with the content. Just choose wisely & follow the path entirely.
- Python, visit the Codecademy website to learn interactively. Also, Datacamp is an option but a paid service. (you can use a 2-month free trial using Visual studio Dev Essentials method, Search online for a video.)
- Check Google for Education | Official learn python website for good documentation on concepts.
- To learn more on programming refer Introduction to computer science & programming using Python via MIT OpenCourseWare.
- R: Visit, Datacamp to start learning interactively.
4. Most importantly: Practice! Practice! Practice!, Yes, you guessed it right; you can get knowledge about technology via course but, working on the projects adds a feather to the cap.
a) Use the various online platform to learn & practice the code, a few of these include. (Not in any specific order)
- Use GeeksforGeeks or W3School for clearing the basics concepts.
- Refer Codeforces & solve problems based on the difficulty level (easy>>medium>>difficult). Also, other websites you can refer to are —
- FreeCodeCamp
- Hackerrank
- HackerEarth
- Leetcode
- Codechef.
Remember not to indulge in multiple websites, choose as per your understanding & style of learning from the options available & build one strong profile to showcase rather, having an average profile on multiple websites.
b) YouTube channels:
Note: “I am providing some additional resources so, you do not spend more time reading blogs about resources, etc….. So, be very cautious & spend a considerable amount of time in sorting the resources as per your goals but once started with, try to avoid any divergence & stick with the track”- All the best with it.
4. Understand the Databases.
As you progress in the field, you’ll come across many types of data, usually in the form of text (also, images, audio, video etc but let’s discuss that later). So, a database is almost always used to store the data. It can be MySql, Postgres, MongoDB, Cassandra etc.
Knowing any of the databases becomes essential. Now, choosing which one is left to the type of industry, company etc. you want to work with. Some of the key resources to learn these tools are listed below:
Learn SQL
- Mode’s SQL for absolute beginners.
- MongoDB University
- SQLZoo: A series of practical exercises and quizzes for students of SQL.
- DataMonkey
- Datacamp: SQL (Check if the style suits your learning way, always good to have an option)
- Advanced SQL Server
- For Indexing: get into immense details to the concepts.
- Quip: Practice The Best Medium-Hard Data Analyst SQL Interview Questions, here.
- SQL Case studies: Problems to solve, related to real-life data science problems.
- Practise SQL skills on Leetcode | Hackerrank |
5. Master Data Cleaning, Exploratory Data Analysis & BI Tools.
Why? It is not very easy in the real world to fetch clean data. When the data is fed to the Machine Learning algorithms, there are a certain set of rules to be followed since everything beneath is Mathematics. Thus, data munging (in simple data cleaning) is a very important step even before we start with the ML model building.
“ Typically 60% of the time is spent on cleaning & dealing with the information, 19% in gathering the correct information, which means Data Scientists spend around 80% of the time in the aforementioned activities than building an ML model.”
“Garbage-in, garbage-out”
Data Cleaning:
Course:
- Getting and Cleaning Data on Coursera by John Hopkins University.
- Intro to Data Analysis on Udacity.
- Data Wrangling with MongoDB: On how to gather various types of data, essentials of storing data, best practices of data cleaning.
Books:
- Data Wrangling with Python: Tips and Tools to Make Your Life Easier, by Jacqueline Kazil
- “Data Cleaning” by Ihab Ilyas and Xu Chu.
- “Best Practices in Data Cleaning” by Jason Osborne.
Tools:
- Data Wrangler
- dplyr
- OpenRefine
- Python & pandas
Reporting:
It is essential in the business to convey the message to others; the audience may include your colleagues, who know about the domain & some who do not have any idea of the technical aspects. In these cases, the visualizations & interactive reports help a lot in representing the work & make it easier for someone with minimum technical knowledge, understand the business outcome.
Most Widely used BI Tools
Tableau:
One of the popular visualization tools, where you can spend around 3–4 weeks to get a good hold on basic-medium reporting skills.
- Click here to decide on a learning path & visit Tableau free learning videos.
- Tableau e-learning platform. (you can have a year of access on the student version & log-in.)
- Data visualization projects for reference.
- Data Visualization Catalogue for reference.
PowerBI:
Another visualization tool by Microsoft & is quite popular due to its affordable price & features when compared to others.
- Official Microsoft website to learn power BI & Community (Open in the browser if needed) for support.
- Webinar from Marco Russo to dive into good Power BI training resource for visual type learning.
- Microsoft YouTube training.
“There are different visualization tools offered by Qlikview, Datawrapper, Plotly, Fusion charts, etc. to name a few, I mentioned the most widely used. You can always check with the target domain & companies you want to work with & then, decide about any particular tool.”
6. How to get experience?
Given the rise in demand & increased competition, it is quite important that you showcase your work via a strong portfolio & have good projects (Hopefully will write another blog on this)
- If you are an undergraduate/recent graduate, try to get an Internships from Internshala or other platforms related to this field by any means.
- Contribute to the Open Source projects on GitHub.
- Participate in competitions & hackathons on
- Kaggle
- HackerEarth
- Machine Hack
- TechGig
- OpenML
“Work on a pet project, it very well may be anything within your present work where you are attempting to tackle some issue and legitimize how utilizing ML is more useful to the business outcome. I can’t emphasise more on how important it is in the entire Data Science/ML career to prove the business impact of a solution in both tangible & intangible way”
A Few ways you can improve your profile:
- Accomplish Bootcamps if time permits. Dphi is one of the good platforms to start from.
- Showcase these projects, code, articles, blogs, etc. on websites like GitHub | Kaggle | Tableau Public.
- Create a Strong Profile on Coding platforms in any of the above-mentioned lists.
- Develop at least 2–3 end-end Machine learning projects from model building to deployment (if possible you can start from data acquisition) to have a good understanding of complete pipeline.
- Refer to one of the sample projects on my GitHub for an end-end ML project.
7. What & Whom to follow?
For Blogs & Articles.
- KDnuggets
- TowardsDataScience
- MachineLearningmastery
- Machinelearningknowledge.ai
- Analytics Vidhya
- Analyticsindiamag
- Datamuni
- MadewithML
YouTube Channels to follow.
- 3Blue1Brown: For intuitively understanding mathematics.
- Applied AI Course: Browse through the free available videos to understand the topics in-depth & also good career advice provided.
- MIT OpenCourseWare: For an in-depth understanding of concepts
- Lex Fridman: Videos exploring research topics in artificial intelligence, deep learning, autonomous vehicles, and beyond.
- StatQuest with Josh Starmer: For statistics & understanding ML topics intuitively & of course for his intro songs too :)
- Deeplearning.ai: For deep learning concept. (have it for future reference)
For paid courses with proper structures.
Not in any particular order & not representing any of it. There is already all the content available out for no cost but prioritizing, filtering the contents is a great challenge & is what up to you, thus, this article to help you.
- Applied Machine Learning Course: One of the best-structured courses available.
- iNeuron Academy.
- 365 Data Science.
People to Follow via Twitter or other platforms
“in no particular order”
- Andrew Ng
- Andrej Karpathy
- Kai-Fu Lee
- Machine Learning Mastery
- Andreas Mueller
- Monica Rogati
- Sebastian Raschka
- Fei-Fei Li
- KirkBorne
- Yann LeCun
Job Searching Websites to follow (in no particular order)
Most of the job you’ll find is via a proper network & referral so, build a strong LinkedIn profile & connect to relevant people & share your thoughts & knowledge via articles, posts, projects etc. Improve your communication: Verbal & written. Approach connections politely to learn more about the industry. Hope you all do great in your career & wish you the best of luck. Make sure to use a job alert feature to get updates.
- LinkedIn Jobs
- iim jobs
- AngelList
- Cutshort
- Indeed
- Glassdoor Get the review on companies from employees & also can apply for jobs If any.
- Snaphunt
- Hirist
- Talent.com (jobs from multiple websites)
Resume Tips & Tricks
- Referrals, connections, colleagues are the best way to get the maximum benefit, the majority of the jobs are not posted online & if so, there are thousands of candidate & a Bot to fight with. Try your level best to leverage your network else, you can follow a few tips as below.
- Thumb rule: Keep it simple, to the point & most important, to 1 page, unless you have 10+ year experience.
- Always make your resume ATS friendly if applying to the job portals.
- Always describe your projects/work using the STAR method (I’ll leave this to your research.)
- Avoid using hyperlinks unless sending straight to someone’s inbox.
- Avoid using tables, headers, footers, lines etc. as these may not parse the information correctly.
- Always tweak resume to the job description. I know it is quite overwhelming to do it all time, but target companies & roles & approach people via LinkedIn.
- Match your resume to the job description & target at-least 70% plus match, (though if it is 50% it’s considered good), but the more your match percentage the more likelihood of your profile is shortlisted. Sharing a tool “Jobscan” use this to learn about the practice & it has some good blogs & resources to refer to.
“Hope this helps you in getting closer to your goal, help you progress in career”
For Interview Preparation/Practise
- How to answer common Interview questions: Click Here to access the tips from Harvard Business School Online.
- The Phone Interview Cheat Sheet: Tips for Success: to know more on, How to handle an interview via phone. Quite important to learn post-pandemic. Sheet-1: Click here, Sheet-2: Click here
- Test your knowledge & understanding of the domain by giving an exam on Workera website
I highly recommend checking this roadmap, One of the best roadmaps for Machine Learning by- Daniel Bourke, Click Here for the access.
Hope, you all liked the content & have some idea, to begin with. Would like to thank the entire community of this field as, without their help, it wouldn’t have been possible to learn & come this far. Kudos to each of you working in this field & sharing knowledge.
I’ll be sharing more content related to machine learning/Data Science/Data Analysis, covering them in detail & hopefully on advanced topics. Till then, Take care & be safe.
IF YOU LIKE IT, PLEASE SHARE IT WITHIN YOUR NETWORK SO, ALL BENEFIT FROM THIS & SAVE SOME VALUABLE TIME WHILE SEARCHING FOR RESOURCES.
Would love to get feedback from you all & welcome suggestions or other resources you find beneficial. please, feel free to share your thoughts in the comment section below.
“Will be sharing the same on other similar websites To help a wider community”
Combined Resources
Click Here to access all the above-mentioned resources in a Notepad for reference. (You have the access to edit the file & a backup link here.)
References:
- https://analyticsindiamag.com/top-10-free-resources-to-learn-microsoft-power-bi/
- https://towardsdatascience.com/how-id-learn-data-science-if-i-could-start-over-2-years-in-b821d8a4876c
- Get some insights by listening to videos from Applied AI Course YouTube channel from Srikanth sir, Cofounder.
- https://analyticsindiamag.com/top-10-free-resources-to-learn-microsoft-power-bi/
- https://analyticsindiamag.com/5-popular-hackathon-platforms-data-scientists-should-know-about/
- https://medium.springboard.com/30-twitter-influencers-you-have-to-follow-for-ai-machine-learning-977587b6406e