Data Science is like Cooking

Before diving into the specifics of machine learning and data science, let’s take a high-level look by comparing it to cooking.

Traditional Programming vs. Machine Learning

Let’s first get acquainted with the difference between programming and machine learning. At the surface, both may seem similar and so it would be a great idea to distinguish between the two.

In his phenomenal video on 2020 Machine Learning Roadmap, Daniel Bourke (subscribe to his YouTube channel) made an analogy on the cooking of roast chicken in his comparison of traditional programming versus machine learning.

He starts by explaining that in traditional programming, we start with the inputs (a box of vegetables and raw chicken) and the explicit programming steps (the cooking recipe), which will ultimately result in the cooked roast chicken. Note here that the recipe (analogously the code) dictates exactly what to do.

As for machine learning, we start with the inputs (a box of vegetables and raw chicken) and the desired output (the cooked roast chicken) then allow machine learning to figure out the cooking recipe (the learned decision rules derived from the trained model) on its own.

Here, we can see differences between these 2 approaches where the former (programming) required all details to be specified in order to cook the chicken while the latter (machine learning) allows the model to figure out on its own the steps to take.

Cooking vs. Data Science

Continuing with the metaphorical analogy of data science with cooking (check out Ken Jee’s awesome YouTube video and Medium article on this topic) we can further see many resemblances between the two topics.

In Ken’s video/article, he starts by comparing data scientists to being chefs. The purchase of ingredients is compared to the process of data collection. Next, the cleaning of vegetables is equated to data cleaning. Then, cutting, slicing and dicing is compared to feature engineering. The heating temperature and cooking time is compared to hyperparameter optimization. And finally the customer experience is compared to the deployment of the model as an application programming interface (API) endpoint.

The Cookbook

Central to the success of any cuisine are recipes and cookbooks (or the knowledge contained hereof) that chefs uses to craft dishes. Likewise, cookbooks also exists in programming and data science, which are in fact referred to as cookbooks such as Deep Learning Cookbook, Machine Learning with Python Cookbook, R Cookbook, etc. In addition to cookbooks, API documentation serve as indispensable tools that data scientists use in coding their machine learning workflow.

The Art of Cooking (and Data Science)

So what does it take to grow as a data scientist? Aspiring chefs undergo intensive training, workshops, school and certifications (so do data scientists). Chefs create new cuisines through experimentation (so do data scientists). Chefs compete at various cooking competitions (so do data scientists, think Kaggle). As we can see, it takes much more than a formal education to become a chef and data scientist. But this is only the beginning, as the learning never stop as it may take a lifetime to perfect the art of cooking (and the art of data science).

There are ample resources for data scientists to learn and grow and this is listed below:

• YouTube — Data Professor, Ken Jee, Krish Naik, codebasics, Daniel Bourke, Tina Huang, StatQuest, etc.

• Medium — Chanin Nantasenamat, Ken Jee, Khuyen Tran, Richmond Alake, SeattleDataGuy, Rebecca Vickery, Terence Shin, etc.

• Books — Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow, R for Data Science, Build a Career in Data Science, etc.

• Online Courses and Certificates—365 Data Science, DataCamp, Coursera

For more information on how to become a data scientist or on staying up-to-date as a data scientist, you might be interested in the article that I wrote Learn Data Science in 10 Steps.