Hacker News new | past | comments | ask | show | jobs | submit login
How to learn Deep Learning in 6 months (towardsdatascience.com)
89 points by antonp on Feb 6, 2018 | hide | past | favorite | 10 comments

Here's Peter Norvig arguing that you shouldn't:

Teach Yourself Programming in Ten Years


The article is about programming, specifically about books etc advising you on how to learn programming in ten days etc, but it totally applies to Deep Learnign also.

Deep Learning is an enormous discipline. In the current point in the hype cycle, there are literally thousands of papers published every month. Personally, I doubt anyone, evern the luminaries of the field are keeping up with it. 6 months is about enough to dip your toes into the preliminaries, and that's if you're willing to skip the relative background [1].

Of course, that's all true if you want to learn. If you just want to train yourself in the use of a few tools, then 6 months is plenty. But don't expect to be able to solve any new problems that way- someone else will always have to do all the hard work for you.


[1] A bit of background will help with questions such as: Why was Deep Learning necessary in the first place? (hint: search for "constant error carousel"); did Deep Learning just spring fully-formed from the brow of Hinton/ Schmidhuber? (hint: read wikipedia on "machine learning" and "Artificial Intelligence"); etc. If you're not asking those questions then you're holding it wrong.

I totally agree with you.

I started Deep Learning Specialization in Coursera: https://www.coursera.org/specializations/deep-learning last month and almost finished it, but I realized this field requires a lot of expertise, not something you can learn in a month. What I learned in the courses was just a basic topics in Deep Learning and how to use Numpy, TensorFlow and Keras.

I was considering diving into a Data Science job and started that specialization as a starting point, but I just realized how foolish I am. Chances are I'll find a job, but it definitely takes another 10 years/10,000 hours to master this discipline.

Anyway, the specialization is wonderful and Dr.Ng explains complicated Deep learning topics in a way that is understandable for everyone. So if just learning is what you want, you should take it, but I don't think you are prepared for a real world Data Science job after finishing it.

How many of those papers are actually necessary, or move the field forward, and how many are "publish or perish" noise?

How to make background to a picture that is geeky for geeky subjects in 10 minutes:

1. find javascript file that has been compressed on your system. 2. open in color coded editor 3. take screenshot

there you go, perfect 'code' background for your deep learning article!

I'm trying to do something similar by starting the other way around (first why it works, then how it works) and so far it's working for me pretty well.[1]

For reference, I'm in high school, so I don't really have the necessary math background to understand everything first try-but I think I'll be able to push through and understand it.

The top down (fast.ai) approach didn't really work for me - as I felt like I didn't really understand how/why everything works. Learning the math behind it forces you to understand.

[1] https://nikhiljha.com/2018/02/06/Neural-Networks-First-Steps...

Sounds like a project manager. Step 1) Prepare for project. 1-2 weeks. Step 2) Do programming 3-4 weeks. Step 3) Unit tests and QA 2 weeks.

My impression is that a lot of programmers have already followed a similar path, would like to find a job doing ML, but haven't found a way to make it happen. Does that sound correct? I'm not saying the effort isn't worth it anyway.

How to learn in 3 months:

Complete said material in 3 months.

With double the effort too ie instead of 10-20 budget for 20-40 hours each week.

For some that might be triple the effort. I guess it depends on a lot of things.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact