Online Machine Learning demonstrations that we like
We’ve been doing a lot of research as part of our own journey to deepen our understanding of machine learning techniques during the past 18 months. As we visit more and more clients to explain machine learning to people, what it is and how it can help companies achieve various goals more efficiently, we have collected a number of online demonstrations that have impressed us.
Feel free to share and comment on this post. We think these are very helpful to raise awareness about the possibilities that machine learning can unlock for us all.
Google Vision API
https://cloud.google.com/vision/
This is an example of how access to data is a key factor when it comes to machine learning. Love it or hate it, but Google is one of the companies that probably collects more data about it’s users, their activities and their preferences than any other in the public space. Linking documents to search terms was their biggest triumph and they have expanded this into images and videos. This enables them to train machine algorithms to “see” what any image contains and catalogue it accordingly.
Reinforcement Learning using Tensor Flow
https://github.com/nivwusquorum/tensorflow-deepq
This repository’s owner, Szymon Sidor, was kind enough to share his game played by an algorithm written using Python and Google’s TensorFlow. Something really visual to show how the algorithm is changing its behaviour as it learns.
From photo to Fine Art
http://thenextweb.com/creativity/2015/08/31/machine-or-picasso-this-algorithm-can-transform-photos-into-fine-art/
This one has no online demo as such. It’s just an article, but the images speak for themselves. If anyone finds an online demonstration of this, please let us know! We decided to include this one, because it opens one up to the idea of machines not only learning, but creating new data, or new information. Music, paintings and other artworks produced by computers? See for yourself!
Raspberry Pi robot that tells you what it is looking at
http://diydrones.com/m/blogpost?id=705844%3ABlogPost%3A2316662
The author, Chris Anderson, built a low cost robot and took the time to explain how he did it. His robot uses the Google Tensorflow library to recognise objects in the camera’s view.
Microsoft Azure online age guessing algorithm
http://how-old.net
Here you can upload photos of yourself and probably your colleagues, your boss and your friends after which this algorithm guesses your age.
Text to handwriting algorithm
http://www.cs.toronto.edu/~graves/handwriting.html
This demonstrates how a machine can change its prediction based on prior events or outputs. When writing with a pen we all know how letters would look different depending on which letters precede or follow the letter as you write. In the same way we train ourselves to write in this way, a machine can learn those patterns.
