-1

I want to understand the maths behind Machine-Learning, and also be able to apply it. Today, I thought I'd make a start. So I did 2 things:

1) I read the first few pages of https://www.amazon.co.uk/Deep-Learning-Adaptive-Computation-Machine/dp/0262035618 , the so called 'AI' bible.

2) Completed the first 2 weeks of Coursera course on machine learning https://www.coursera.org/learn/machine-learning offered by standford uni for free.

In both cases, I went over simple Matrix-Vector multiplication and simple linear regression, which felt very dry.

Now I'd like to know if anyone has actually read that book and could comment on the complexity of maths behind it? It feels like I'm falling asleep watching the video tutorials and I would prefer to read at my own pace, studying the book.

GRS
  • 2,495
  • Most of the time it's linear algebra indeed ... Linear/Logistic regressions, PCA, HMM, Kalman/Particle Filters. Calculus and probabilities - expectation maximisation, maximum likelihood, Bayesian inference. That's not the point, the point is how simple maths can do great things. Also, this is a good book https://www.amazon.co.uk/Artificial-Intelligence-International-Version-Approach/dp/0132071487/ Re the course, stay tuned, there will be more material there – rtybase Aug 04 '17 at 23:16
  • @rtybase I've decided to do 1 week of lectures every day for the summer, and read the book too. I'm also worried about what to do next year... I have an offer for mediocre finance jobs which 1) I have no interest doing 2) Don't pay enough. Looking at data science jobs, all want at least a Masters level degree (although I got a disctinction award in my BSc Maths & Stats, it's not enough to land interesting jobs). I wonder if I should just read books and apply for masters in 2018, or what to do in general – GRS Aug 04 '17 at 23:46
  • If you can afford doing Masters, do it. It will be more complicated later (due to job & other commitments). If you plan to do Data Science, read a few books just to know what it is about, but it is essential to learn a programming language (Python or R for example) and a few ready to use frameworks/API's for Machine Learning and data visualisation, most of the time that's what Data Scientist are paid for. – rtybase Aug 05 '17 at 00:03
  • @rtybase Yes, unfortunately I missed the deadline for applying to masters, so I can either work for a year, travel (and read the books) or work for a startup/create one – GRS Aug 05 '17 at 00:35

1 Answers1

1

I consistently find that video lessons/courses are too slow for me to feel like I'm learning at the rate I'd like to be. I prefer to find books or especially online PDFs on the exact topic I'd like to learn about and tackle those, and leave the videos to supplement my understanding.

Reading Pros:

  • Learn at your own pace
  • Feels like there exists a wider range of topics as well as more specific texts on a single topic
  • Tends to have more technical language
  • Less intrusive than video (for reading at a coffee shop for instance)
  • Offers more references/further readings than videos

Reading Cons:

  • Requires more work keeping yourself focused
  • Can start to feel dry after too long of staring at a page
  • Tends to be more difficult to visualize a concept (a large part of my understanding)

Video Pros:

  • Requires less work on your end; hit play and sit back
  • Have real people going through each step with you
  • More options for visualization

Video Cons:

  • Usually much slower than books
  • Often includes wasted time (taking attendance, recap of previous video, stopping to ask questions (usually never good ones), end of the video asks to like/subscribe, etc.)

That said, sometimes I find videos that really are worth watching, for example I remember an old series on complex analysis that is on youtube and I really appreciated, also anything by Tadashi Tokieda like the lectures he gave in South Africa. Other than that though, if I really want to use a video to stipend my book learning, I'll set it to 1.25/1.5x speed which helps to make up for how slow it goes, the biggest drawback to videos in my opinion.

  • Thanks a lot, I share this exact feeling, and it's especially difficult to decide on a learning method before learning a new topic – GRS Aug 05 '17 at 00:30