Really excited about how much clear and concise documentation is available here. I've been considering looking to machine learning and tensorflow to at least get the concepts. This is just what I need to push me over the edge and get started
(You need to set up an account on packtpub.com to participate; in addition to allowing downloads, books are available in one's account to be read online).
The book will be available through Thurs., Aug. 31 till 4PM PDT (i.e. midnight UTC), when the free daily e-book changes.
As someone who utilizes machine learning but doesn't have a great understanding of neural networks, I've not had good experiences with TensorFlow. Myself and a friend attempted to use it for a project but found a lack of examples and the documentation to be very unclear. Even doing basic things for which tutorials did exist presented us with lots of warnings and errors. Debugging one of these, for example, led us to believe that certain functions were incomplete or abandoned, without clear messaging in the documentation that this was he case.
For those interested in diving into machine learning, I'd recommend not starting with neural networks. I think there's a belief amongst those unfamiliar that a neural net will be 10x better than anything else, but this is not always the case, and you can accomplish a lot with simple functionality found in scikit.
Agreed about scikit-learn. Anecdotally I've found that most "deep learning" and "big data" problems I see in industry could be trivially solved using scikit-learn on a reasonably powerful laptop.
Also if you want to really understand neural networks I can recommend Andrew Ng's new deep learning course on Coursea. You start with deriving the underlying math and implementing a simple (but fully functional) neural network using just basic numpy, so that when you get to tensor flow and the latest cutting edge techniques you actually understand what's happening under the hood and how things actually work.
To me the irritating point is all the tutorials with MNIST, but I'm not able to find only one explaining how to prep a good dataset to use with it (the one in section 5 is ok, but I feel it is not really practical?).
Neither one about trying to classify something literary from scratch.
Please stop with MNIST sample and character recognition ones...
Has anyone used TensorFlow for any hobby projects? I would be interested to know what kind and how it worked out. This is a pretty neat tutorial, I'll have to set aside more time to look it over.
9 comments
[ 3.2 ms ] story [ 20.2 ms ] threadhttps://www.packtpub.com/packt/offers/free-learning
(You need to set up an account on packtpub.com to participate; in addition to allowing downloads, books are available in one's account to be read online).
The book will be available through Thurs., Aug. 31 till 4PM PDT (i.e. midnight UTC), when the free daily e-book changes.
For those interested in diving into machine learning, I'd recommend not starting with neural networks. I think there's a belief amongst those unfamiliar that a neural net will be 10x better than anything else, but this is not always the case, and you can accomplish a lot with simple functionality found in scikit.
TLDR, glad to see more tutorials!
Also if you want to really understand neural networks I can recommend Andrew Ng's new deep learning course on Coursea. You start with deriving the underlying math and implementing a simple (but fully functional) neural network using just basic numpy, so that when you get to tensor flow and the latest cutting edge techniques you actually understand what's happening under the hood and how things actually work.
To me the irritating point is all the tutorials with MNIST, but I'm not able to find only one explaining how to prep a good dataset to use with it (the one in section 5 is ok, but I feel it is not really practical?).
Neither one about trying to classify something literary from scratch.
Please stop with MNIST sample and character recognition ones...