Machine Learning. Get excited!

For a soft intro lets talk about regression analysis the un-sexy grandfather of predictive analytics: regression analysis is the statistical process of estimating the relationship between a dependant variable (Y) and the causality on one or more independent variables(x). A good example would be if we wanted to test Y “House Price” we would test this against a combination of x variables such as “Bedrooms”x1, “Bathrooms”x2 etc. So using the diagram we could at some point conclude that houses with more values of x have a larger value of Y.

Cool so now we got that out of the way we can all forecast house prices and will be professional investors in no time.

The most interesting thing about Machine Learning – "it's different".

Forecasting a result through examples: That’s right by example! Where in the first example we look at testing by giving inputs and attempting to find meaning between Y and some x’s that have causality we do not need to worry about this so much anymore. Machine learning builds features based on example inputs….. so what does this mean? The best way to describe this is that machine learning takes the data that it is given and forges a memory that holds some point of reflection which is generally why it takes about 10,000 images to recognise a face, this is a good thing this means now your computer can recognise a face all we need is a whole lot of examples. When you think about it we do this as well, we might see a red woollen jumper, smell some perfume or hear an old song and it stirs a distant memory this is our own mind is very similar to ml classifying something to define it with meaning based on previous examples.

So why get excited about this distinction? Well example based learning is intelligent and far more robust than any forecasting method that is precedent. This is possible due to the features created via the examples the ML is given and this is far more powerful than any human made attempt to construct a manual model of causality. Cool right!

This is very important especially when we look at the dimensional aspects of data. Where the above graph is 2 dimensions and we can take a quick look and see what is going on. ml can take 100's of dimensions that would give a human analyst a nose bleed and produce a highly accurate result without faltering.

Crystal Ball Time: In the not so distant future rather than input software the world should start to see example based programs that can conduct analysis and evaluate environments and give meaningful results. What does this mean? well like anything that isn’t naturally occurring like trees, grass and dirt everything else you lay eyes on is just another good idea added to the human collection of things. Data, People, Things and these things are coming and when they do everyone's lives will be better for it granted the first line of code is always...

skynet = False