>Business >Self-study guide to machine learning

Self-study guide to machine learning

There are various things you can do to get acquainted with the discipline of machine learning. 

There exist assets online such as books and course you can enrol into, contests you can enter into and utilities you can leverage. 

In this blog post by AICorespot, we want to provide some structure to follow for new entrants to the study of machine learning and indicate a loose ordering of what to take on in your progression from programmer to machine learning maestro. 

The four stages of machine learning 

We can adopt the perspective that there are four levels of fluency within machine learning. This model facilitates us to think about what you require and the activities that are available and when a good time to handle them might be. 

The four stages of machine learning proficiency are: 

  1. Beginner 
  2. Novice 
  3. Intermediate 
  4. Advanced 

Beginner is demarcated from novice as it provide motivation to absolute beginners have paths available to them if they choose to get into machine learning. 

We are going to dive into every one of these four stages and look at online resources and activities that can assist somebody at one stage learn more and advance their comprehension and skill levels. 

The breakdown is merely a suggestion, and it is highly probable that some activity or resource at a level prior or after can be very useful and relevant at a provided level in the breakdown. 


A starter, or beginner, can be defined as a programmer who is interested in machine learning. They might have begun to read a book, Wiki page, or have taken a few short courses, but they don’t really get the meat and potatoes of it yet. They’re experiencing frustration as the advice they’re obtaining is for intermediates and sophisticated learner levels. 

Starters are in requirement of a gentle introduction. They need to shift away from code and textbooks and courses. They require the whys and whats and hows indicated to them first to establish the foundation for novice-level material. 

Some actions and assets for the absolute beginner are: 

  • Introductions from books: Get into introductions to good machine learning and data mining books for programmers, like: 
    • Machine learning for Hackers 
    • Programming Collective Intelligence 
    • Data Mining: Practical Machine Learning Tools and Techniques 
  • Overview videos: Look into presentations that provide a summarization of machine learning to laymen learners. Some instances consist of: Interview with Tom Mitchell and Peter Norvig on big data Facebook Tech Talk. 
  • Talk to individuals: Question them on how they begun in the domain, what assets and resources they recommend for starters, what thrills them about the domain. 


A novice is somebody who has had some prior exposure to the domain of machine learning. They have consumed a book or have completed a course. They have now been bitten by the bug and they desire to know more. They are beginning to get it and want to start to get things done.  

Novices need things to do. They require to be put into action to have the material grounded and integrated in current know-how structures such as the programming languages they are aware of or the problems they are accustomed to finding solutions to. 

Some activities and resources for novices are: 

  • Finish a course: Enlist in and finish a course like the Stanford Machine Learning course. Jot down a lot of notes, finish the homework if possible, and ask a lot of queries. 
  • Get into some books: Not dry textbooks, but approachable books such as those listed prior aimed at fledgling programmers. 
  • Learn a tool: Learn to drive a tool or library like Scikit-learn, WEKA, R, or similar. Particularly, learn how to leverage an algorithm you have read or learned about in a book or course. See how it functions practically and get accustomed to attempting things out as you get to know them. 
  • Author some code: Implement a simpler algorithm such as a perceptron, k-nearest neighbour or linear regression. Author small programs to demystify strategies and methods and learn all the micro-decision needed to make it function. 
  • Finish tutorials: Follow and finish tutorials. Begin developing up a directory of small projects that you have finished with datasets, scripts and even source code you can fall back on, read and think about. 


A novice has studied some literature and finished some courses. They are aware of how to drive some tools and have authored a bunch of coding both implementing simple algorithms and finishing guides. An intermediate is starting out on their own, developing their proprietary projects to learn new strategies and interacting and learning from the larger community. 

The intermediate is becoming aware of how to go about implementing and wielding algorithms precisely, in a competent and robust fashion. They are also developing the skills and abilities of spending a lot of time with data up front, cleaning, summarizing, and contemplating about the variants of questions that it can answer. 

Some activities and assets for the intermediate are: 

  • Small projects: Develop small programming projects and experiments where machine learning can be leveraged to find a solution to a problem. This is like designing and carrying out your proprietary tutorials in order to explore a strategy you’re fascinated in. You might implement an algorithm or link to a library that furnishes the algorithm.  
  • Data analysis: Get accustomed to exploring and summarizing datasets. Go about automating reports, be aware of which tools and utilities to leverage when, and look for information you can look into, clean, and on which you can go about practicing strategies and techniques and communicate something interesting.  
  • Read textbooks: Read and digest textbooks on machine learning. This might very well need skills to grok mathematical descriptions of strategies and techniques and acknowledging formalisms that detail classes of problems and algorithms. 
  • Author plugins: Author plugins and packages for open-source machine learning platforms and libraries. This is an exercise in learning how to author robust and production level algorithm implementations. Author your own plugins on projects, query for code reviews from the community and work to get the code integrated into the platform if possible. Obtaining feedback and learning is the objective. 
  • Competitions: Take part in ML competitions, like those associated with conferences or offered on platforms such as Kaggle. Take part in discussions, query other people with doubts you might have, get to know other practitioners are tackling the problem. Include to your repository of projects, strategies, methods, and code from which you can draw. 


A sophisticated practitioner has authored a ton of code either integrating ML algorithms or implementing algorithms themselves. They might have taken part in contests or authored plugins. They have gone through the textbooks, finished the courses, and have a wide comprehension of the domain, as well as an in-depth know-how on a few critical strategies of which they prefer. 

The sophisticated practitioner develops, deploys, and maintains production systems that leverage machine learning. They keep themselves updated on new developments in the domain and eagerly search out and educate themselves on the nuances of a strategy or method and tips passed around from other frontline practitioners such as themselves.  

A few activities and assets for the advanced practitioner are: 

  • Customizing algorithms: Alter algorithms to meet their requirements, which might consist of implementation of customizations outlined in conference and journal papers for similar problem domains. 
  • New algorithms: Developing completely novel strategies and methods on the basis of the underlying formalisms to meet the challenges they face. It is more about obtaining the best outcomes possible instead of advancing the frontier of the field. 
  • Case studies: Read and even recreate case studies finished for machine learning contests and by other practitioners. Those “how I di it” and posts are typically loaded with subtle professional tips for data prep, feature engineering and technique usage. 
  • Methodology: Systemization of processes, whether formally or for themselves. They have a method to tackle problems and obtain outcomes at this juncture and they are actively seeking ways to further refine and enhance that process with tips, best practices, and new and improved strategies. 
  • Research: Going to conferences, reading research papers and monographs, having convos with specialists in the domain. They might author up some of their work and submit it for publishing, or just drop it in a blog post and go back to work. 

Mastery is an ongoing journey, and the learning process never reaches a conclusion. One could stop and take a detour at any juncture along this journey and turn into the “competition guy” or the “pro library guy”. As a matter of fact, we expect that such detours to be the norm. 

This breakdown could be interpreted as a linear path of the technicians journey from starter to advanced level, it’s intentionally programmer centric. This article is merely a listing of AICorespot’s suggestions of the varieties of activities to manage if you identify yourself hungering for a more at a particular level. 

Add Comment