Machine Learning for Beginners

[K D]

I'm looking for some learning resources to get started on Machine Learning. Can anyone give some pointers?

 

[mackle]

I started taking the ML course on Coursera. Enjoyed it but then life got too busy.

 

[BackupProphet]

If you have coding skills, getting started with machine learning is not that difficult.

Safari Online books has a nice video tutorial on deep learning with Python, Keras(Tensorflow)
Deep Learning with TensorFlow: Applications of Deep Neural Networks to Machine Learning Tasks
I think for many, that is a good start.

Also be careful about what library you start with, Scikit-Learn for Python cover almost everything related to machine learning except the deep learning part. For Deep Learning, Keras is a good start for almost everyone. Tensorflow, Theano, PyTorch are for the experts.

 

[tonydsouza]

To start with machine learning, you should understand the basics. Recently I had gone through a resource which was easily understandable and was in the form of infograghic which is very interesting and engaging.
Machine Learning- Deciphering the most Disruptive Innovation : INFOGRAPHIC | Fingent Blog

 

[vv111y]

Good SubReddit. Great for asking questions and guides.
Learn Machine Learning

Their wiki has links to other guides, I haven't looked at them, but I have found sub-reddits suggestions are good
index - learnmachinelearning

Another subreddit
Machine Learning Questions
Machine Learning Questions

Mathematical Monk.
This digs into the math so you have a good understanding of the fundamentals. Done like Kahn Academy.
Machine Learning Playlist - YouTube

resource list:
ujjwalkarn/Machine-Learning-Tutorials: machine learning and deep learning tutorials, articles and other resources
ujjwalkarn/Machine-Learning-Tutorials

EDIT:
oh, and there is
fast.ai · Making neural nets uncool again
fast.ai · Making neural nets uncool again

which I have heard is very popular and gets you up to speed quickly. They do group studies with live coding and online discussions. I plan to check it out sometime soon

 

[MiniKnight]

My issue is going from guide and simple tutorial to actionable. Maybe it's more precise to say getting enough clean data and coming up with how to actually go from data to a model.

That's what's missing in many videos. They'll show you how to do simple, but they're not showing the way to get practical. And I'd prefer to learn through directly solving problems rather than learning how to do text lstms after doing image recognition.

 

[vv111y]

My issue is going from guide and simple tutorial to actionable. Maybe it's more precise to say getting enough clean data and coming up with how to actually go from data to a model.

That's what's missing in many videos. They'll show you how to do simple, but they're not showing the way to get practical. And I'd prefer to learn through directly solving problems rather than learning how to do text lstms after doing image recognition.

Probably fast.ai would be what you want.

You can also use .ipynb's on github for starting points. The code is right there, and some of these notebooks are not toy problems - well, I guess depends on what qualifies as toy. ie. CIFAR? CelebA?
You can also just use the same code for bigger datasets, but then the issue is the hardware.