Where will AI go?

Today I just wanted to share a news article that I have found on where things like Neural Networks will go in the future, could it be marking exams?.

Keras

I am hoping to get back to some core posts on automating office with python soon, however at the moment I am exploring machine learning and neural networks as I personally have some situations where they would be useful to implement, at work I have some use cases and lets face it they are cool. 

In order to improve my home portfolio of machine learning and data science techniques I have been looking into Keras and using it to build a neural network. My hope is that over the next week I can replace the logistic regression model I build for predicting the odds of Dupuytren's patient developing Ledderhose with a Neural network. 

Starting out with Python and Neural Networks:

Ok so when you are first looking into Neural Networks for python you start hearing words like Keras, Tensorflow and Theano.

What is Keras?

Well Keras is a library that sits on top of other machine learning libraries such Tensorflow and Theano. To quote the Wikipedia page "Keras was conceived to be an interface rather than an end-to-end machine-learning framework. It presents a higher-level, more intuitive set of abstractions that make it easy to configure neural networks regardless of the backend scientific computing library."

The Keras homepage actually has a get started in 30 seconds section and although not with my dataset but a sample one I did get a model up and running in a sleep deprived state. It really is simple, I also found the video here.

So Keras does lots of clever stuff to make it easier and many less lines of code to run process in Tensorflow. In fact above is the output given when you kick off Keras.

I have only really give a brief intro in this post however I am hoping have a great post ready soon with my first publishable neural network in Keras.

Neural Networks

So I am not doing a proper post today because I am busy brushing up on my Neural Networks so that I can do some more details posts on this and hopefully how they can be implemented for many different uses. I have a couple of use cases at work that I will be able to use this for. 

The first is something which I logistic regression model could be used for but the new course that I am taking on Deep Learning has a model for detection of fraud using neural networks. The other is for the prediction of customer churn rate. 

The course can be found on Udemy, I am using this in conjunction with the Machine learning hands on book to increase my knowledge on neural networks and machine learning.