Physicist to Programmer: bottom-up vs. top-down learning

Russell Pollari
HackerNoon.com
Published in
3 min readApr 22, 2019

--

Two years ago I left a PhD in physics, joined a startup, and taught myself to be a software developer. I had to radically alter the way I learned and approached problems.

An education in physics is very much a bottom-up approach. Before you can grasp the intricacies of electromagnetism, quantum mechanics, and general relativity, you need a firm footing in the fundamentals. I believe this is the right learning model for physics. If you’re trying to determine the underlying truths about the universe, you should have a solid understanding of first-principles.

This mindset, however, can lead to a “Physicist’s ego”: A sense of superiority over those who don’t seek bottom-up understandings. A distaste for black box solutions. As a physics student, I experienced this “ego” in many ways. Contempt for the “plug and chug” problem solving that engineers employed (using equations but skipping their derivations), refusal to use third-party code or out-of-the-box solutions in my research (considering it akin to cheating), and a general opinion that physics was the superior discipline (except, perhaps, for math).

Then I joined a startup as a full-stack software developer with almost zero programming experience. Turns out all those Matlab scripts I wrote for my research did not even come close to preparing me for the life of a software developer. Early-stage startups have to move fast, so I had to learn faster. The bottom-up approach was no longer sustainable or appropriate.

A full understanding of computer science and programming does require a deep level of knowledge. But, depending on your goals, the level of understanding you actually require varies greatly. Modern programming languages consist of layers of abstraction built on layers of abstraction. They are designed to reduce the required background knowledge of the programmer. You should embrace this.

The best way to learn to code is to start coding. This is cliche advice but it’s important to internalize. Do not let incomplete knowledge stop you from starting. Get comfortable using tools before you fully understand them. Your learning style should be top-down. Work down the layers of abstraction only when you need to.

The physicist in me still recoils at this mindset. But, as an approach to self learning, it’s been very effective. Start learning by pattern matching. How did someone else solve a similar problem? How can I modify their code to make it do what I want to? Eventually, the problems you’re solving will become unique or you’ll hit a bug you don’t understand and you’ll be forced to dive a little deeper in your understanding . You learn to code by coding, but the real learning comes from breaking things and fixing them.

To become a programmer from scratch follow these steps:

  1. Become a copycat: Mimic more experienced programmers
  2. Stay shallow: Learn only what you need to know for your specific problem
  3. Launch early: Start by getting something (anything!) working as quickly as you can
  4. Every bug is a lesson: When you hit a bug, do a deep dive to understand it as fully as you can

Russell is a data science mentor and CTO at SharpestMinds, follow him on twitter @russ_poll

Originally published at http://russellpollari.com on April 22, 2019.

--

--

HackerNoon.com
HackerNoon.com

Published in HackerNoon.com

Elijah McClain, George Floyd, Eric Garner, Breonna Taylor, Ahmaud Arbery, Michael Brown, Oscar Grant, Atatiana Jefferson, Tamir Rice, Bettie Jones, Botham Jean

Russell Pollari
Russell Pollari

Written by Russell Pollari

Machine learning and prog metal enthusiast. CEO at SharpestMinds.com—mentorships for data scientists.

No responses yet