Learning ML or Learning About Learning ML?
Pragmatism versus (over-)planning
We’ve all been there. Our browsers are full of them, our notes are overflowing, and we often have detailed plan on tackling them: online courses about Machine Learning, articles about Machine Learning, videos about Machine Learning.
Roaming around the internet, we are endlessly populating our list of “cool ML resources.” When we find a pre-made curriculum about how we can learn ML in X days, it’s as if we have hit the jackpot. “That’s exactly what I needed,” we tell ourselves, “that’s how I can learn ML.” We are passionate about ML, and we quickly get geared up by shiny new things, courses, learning material.
When we find a resource that resonates with us, we readily envision ourselves sitting behind our desks, with books (or screens) all around us. We see ourselves happily reading page after page, coding line after line, implementing challenging algorithms. We can almost feel ourselves being X days further into the future.
Then we find another article about learning ML, complete with recommended courses, course orders, alternative paths. Another daydream starts.
It happens to anybody at any stage of the ML journey; it recently happened to me. Five years into my ML journey, I realized I needed to re-engage with the fundamentals. I searched for the best courses teaching the math behind ML, looking for resources that provided exercise materials. I compared platform A against B against C. Whenever an offering, paid or free, did not check all (unconsciously required) boxes, I searched on.
When I found a good course, I did one lesson, found that it was not a 100% fit, doubted my decision, and started searching again…
Stop, I eventually told myself, and settled with a freely available resource — I had been caught in my head, escaping the less romantic (but more realistic) day-to-day ML practicing.
As with any other field, it’s easier to mentally visualize ourselves learning ML than actually doing the learning in that moment. And then there’s the point that learning ML is hard. It requires focus, practice, and patience. It’s less straining to learn about learning ML, instead of actually learning ML.
And yes, the internet is full of resources — which is why we have ever-growing lists of ML online courses, blog posts, and videos. But at some point, you need to actually jump to move forward. Preparing the jump is all fine, but preparation without execution is just wasted time.
To make progress in your Machine Learning journey, you sometimes need to be pragmatic. Ask yourself what you want to learn and then take the actual step and sit down to learn. To make that execution easier, use the following approach:
- Quickly pick one resource and stick to it: Strongly limit your research time. Remember that adding yet-another ML course to your list is not progress. Depending on your prior experience, one to two hours of thorough research suffice.
Then, commit to a single resource, whether it’s an online course, a book, or tutorial, and follow it through. Avoid jumping from one shiny resource to another. - Set realistic goals: Break your learning into manageable chunks. Aim to complete a chapter, a coding exercise, or a video lesson in a set amount of time. Daily (or weekly) goals work best because they can be easily grasped.
- Practice actively: Don’t just passively consume content. Apply what you’re learning by solving exercise problems, writing code (my favourite), or working on your own projects.
- Create a routine: Dedicate specific hours to ML learning. Depending on your remaining obligations, two hours first thing in the morning work best. Importantly, use these pre-allocated time slots to continue with your chosen material, not for searching for new stuff.
- Embrace imperfection: Your first few projects, lines of code, or exercises might be messy, and that is okay. If they were perfect, you didn’t need to learn anything new in the first place.
- Be pragmatic: Don’t overthink. Do the immediately required next step and adapt afterwards.
Your ML journey will have ups and downs, it rarely is linear. If you catch yourself over-planning, remember: any action is often better than a never-executed perfect plan.
Learning ML or Learning About Learning ML? was originally published in Towards Data Science on Medium, where people are continuing the conversation by highlighting and responding to this story.
from Datascience in Towards Data Science on Medium https://ift.tt/lfxyzjL
via IFTTT
