Machine learning curriculum: From Rookie to Mastery
We are living in the age of AI. It is almost a cliché to say this at this point. Machine learning coupled with other technologies has the power to change the world, for better or for worse. I for one, champion its cause for making our world a better place.
The beauty of rapid development in this field is that the tools are accessible to everyone. There are plenty of resources online that help you learn and in this article, I would like to put together a list of such resources that will help you navigate your journey. The curriculum suggested here is for a complete novice, assuming you have no background in programming. Depending on your prior experience and expertise, you can skip some of the early stages listed here.
Stage 1: Learning the language
Two languages are primarily used for data science and machine learning — Python and R. Hence we designed the curriculum along these two language tracks
Python Track
Although python is not the only language for machine learning, it is surely the popular choice amongst data scientists, machine learning engineers, and it also comes with a wealth of libraries for machine learning. Most machine learning and deep learning educational courses are primarily taught in python.
R Track
R was originally designed by statisticians for doing statistical analysis, and it remains the programming choice of most statisticians today. It was designed specifically with data manipulation and analysis in mind. Because of that, learning the core skills of data science — data manipulation, data visualization, and machine learning — can actually be easier in R once you’ve gotten through the basic fundamentals.
Note : Unless you have a prior experience working with R language, I would strongly recommend you choose the Python track. R does come with a lot of advantages of its own, but python is easier to learn and you will find a wealth of resources to learn machine learning with it and also powerful machine libraries such as scikit-learn, numpy, pandas, nltk and matplotlib
Stage 2: Data Analysis and Visualization
Every machine learning problem starts with data analysis and exploration. Data visualization is particularly key to identifying patterns, distributions, and outliers in the dataset that you are working with.
Python Track
2. Introduction to data analysis | Udacity | Free
R Track
2. Here you have two options to go for —
Data scientist with R career track | Data camp | Free
Data Analysis and Visualization | Udacity | Free
Stage 3: Get your hands dirty
At this point, take a break, look back at what you learned. This would be a great time for you to try your hand at some data analysis on open data. A great place to start with this would be — Kaggle. Explore some of the datasets available here, pick any one of them and go wild. You can also refer to the work done with these datasets by other users for some inspiration — Kaggle notebooks.
Kaggle is an online community for data science and machine learning practitioners. It hosts some exciting competitions in these fields and offers easy access to some open data. This is a great place for you to practice your learnings so far and also later in your curriculum.
Stage 4: Machine learning basics
Python Track
3. For a gentle introduction to machine learning, while also learning a few tricks of trading, you can try the Machine learning for trading | Udacity | Free. It particularly helped me understand how different techniques of machine learning can be applied to an unrelated field with some domain knowledge.
Of course, this course isn’t mandatory for you to follow, you could also do the course 3 in the R track as an alternative. It is only meant to ease you into the machine learning universe without being intimidated by the jargon.
4. For a broader introduction to the machine learning concepts and workflow, you should next checkout — Machine learning | Coursera | Free
R Track
3. For a general introduction to Machine learning I suggest — Machine Learning for everyone | Datacamp | Free. This is a “no-coding course” and gives you a general introduction to the machine learning universe.
4. For machine learning with R, data camp offers a free career track that covers most fundamentals in machine learning — Machine learning scientist with R | Datacamp | Free.
I highly recommend you do the course offered by Andrew Ng next, to get an in-depth understanding of machine learning concepts. Although Andrew Ng’s course is taught in Python, the intuitions on various topics is universal. The implementations can be translated to R with some effort.
Stage 5: Deep Learning
Python Tracks
There are two deep learning courses that are taught in python and are of great repute — one is by deeplearning.ai and the other is by fastAI. deeplearning.ai’s course is available on courseera and it teaches deep learning with TensorFlow, while fastAI’s course is taught with the fastAI python library, which is a wrapper over the PyTorch library for deep learning. Hence we split the python track into two branches, one with deeplearning.ai, TensorFlow and the other with fastAI
Both PyTorch and TensorFlow are popular open source platforms for machine learning. Other examples deep learning libraries built for python include, Keras, Theano and more.
5. fast.ai | Free
Fast AI is a great introductory course to deep learning. The instructor Jeremy does a great job at explaining complex concepts with much ease. He introduces you to state of the art deep learning model. Most of the implementation is done using the FastAI library, which like I mentioned is a wrapper over the PyTorch library. It is often suggested to learn to build deep learning projects with fastAI library first and eventually transition to PyTorch
6. Next, learn PyTorch— Intro to Deep Learning with PyTorch | Udacity | Free.
Branch II
5. deeplearning.ai | coursera | €44 / month
6. Intro to TensorFlow for deep learning| Udacity | Free
Although PyTorch is gaining popularity in recent years, TensorFlow is widely used in industry and is a very mature library that comes with strong visualisation capabilities and several options to use for high-level model development.
You could choose either of these branch for a deep learning specialisation, or do both for a well grounded understanding of deep learning, as Michael Li summarised in his post here —
R Track
Keras is a high-level deep package first introduced for Python. R offers an interface for Keras that allows you to implement deep learning models with ease.
5. Deep learning with R in motion — Video series that compliments the Deep learning with R book.
6. Once you are done with the introductory course in step 5, I strongly recommend you do either the fast.ai’s or deeplearning.ai’s course on deep learning for in-depth understanding and hands-on experience. Make an effort to translate your learnings from these courses into R with keras support.
Other resources — Books
- Hands on computer vision with TensorFlow 2
- Python Machine Learning
- Deep learning with R
- Hands-On Machine Learning with Scikit-Learn and TensorFlow
- Also checkout the annual subscription for O’Reilly to get access to some great learning content and books
Quick reference guides
- Machine Learning | Stanford | Cheat sheet
- Deep Learning | Stanford | Cheat sheet
- Also our page — ML Cheat Sheet | Medium
- Daniel Bourke’s blog — Starting with no knowledge of programming, he self designed a curriculum for machine learning and mastered it. It is a great source of inspiration for everyone starting on this journey on their own as well.
There are plenty of resources available online for you to learn machine learning. This curriculum is intended for you to clear the clutter.
Feel free to make changes and replace any of the courses listed above, with your own preferences, while ensuring you cover all the 5 stages for a holistic learning.
Follow me on LinkedIn.
You may also reach out to me via sowmyayellapragada@gmail.com
