Searching for tutorials and software about Deep Learning and Neural Nets? Be sure to look at my Resource Page!
Looking for Octave? Go to my Easy Octave on Mac page!

Monday, November 16, 2015

Conference papers are the natural prey of the iPad Pro.

Yesterday I went through the Mandelbrot Tensorflow tutorial.

Today, I went out to get an iPad Pro. It makes a wonderful PDF reader. I find PDFs are really hard to read on a laptop or small iPad,  but the iPad Pro and the Surface Pro have nice screens and can use pens for annotation. 

PDF Expert from Readdle is free this week, and it's my solution to the annotation problem on the iPad Pro. Of course if you don't need to annotate, you can just iBooks to ingest the PDFs and file them away - the trick is to click at the top right of Safari after the download finishes, and choose "open in iBooks".  However PDF Expert does have one trump card: It can index and search through your whole PDF library! 

Others may prefer to use OneNote on the Surface Pro: Here the trick is to run a book or paper through the "Print to OneNote" device, at which point you can drop it into your notebook and use all of OneNote's tools on the text, including and especially search and annotation.  In theory this allows you to quickly search a whole library of saved papers. 

The Apple Pencil is hard to find, but I have the one from 53, it works great, and I love the soft tip when using it as a typing and selection tool. Here is the Amazon link, they also have them at Apple retail stores. By the way, the Paper app from 53 is really really great; it was a great art app before, now it's a great notetaking app. 

By the way, I would be willing to bet that the electronics in the Apple Pencil and the 53 pencil are very similar. 

The improvements in stylus reactivity on the iPad Pro are spectacular even with a dumb stylus, and seem mostly due to a scan rate speedup in the display's touch sensor system. This means that actually any capacitative-display dumb stylus should work well on the iPad Pro.

Certainly the cheap stylus I also tested allowed me to write and draw fluidly with Paper.

If you are looking for test PDFs, there is always the TensorFlow white paper which describes Google's Artificial Neural Network simulation system in design, architecture, and implementation.  In addition to the ANN-specific stuff, it has a lot of abstruse details on the implementation issues for a networked dataflow simulator, and presumably points to references on the dataflow concept which might be worth chasing up. If Google have their way, the dataflow concept is due for a comeback.

The convincing argument which makes it worth learning TensorFlow is that for Google it serves both as a research environment, training software and production tool. This is software which has been tested, has seen production , and will not turn into abandoware come the morning. 

No comments:

Post a Comment

Hey, let me know what you think of my blog, and what material I should add!