Machine learning is becoming quite a fashionable phrase, but what does it mean?  And what does it mean for ordinary people who work at the day job.

You’ll remember the term Artificial Intelligence from back in the 90’s.  You may have come across the term Neural Networks.  In this blog you’ll find lots of comment about Bayesian Artificial Neural Networks – ours.  These are all synonyms for Machine Learning.

It’s actually all mathematics, and its about recognising patterns andcalculating probabilities of what’s going to happen.  It’s really no different to the way we all learn – remembering what happened, so we can anticipate what comes next.

With machine learning we use computers and mathematics in place of our own intelligence.  Hence the term Artificial Intelligence.  This lets usanalyse millions of pieces of information and carry out billions of calculations.  In the process the machine learns.  It stores the results of those calculations and that’s where the intelligence is.

We can  take a simple example of a motor car, driven on occasion by three people – my wife, my son, and me.  My wife drives carefully, never racing the engine, braking smoothy and rarely exceeding the speed limit.  On the other hand my son is the opposite.  He likes to get going.  And then my driving style is different – somewhere in between the two.

Over time the machine in the car learns something about our driving styles and is able to make predictions about how many miles are left in the tank, depending on who’s driving.  It recognises that from patterns of acceleration, braking and speed.

Going a little deeper the machine can relate actual fuel consumption to the driver it’s identified.  If the driving style is consistent but fuel consumption is up, it might recognise the patterns from before, when the tyres needed inflating.  A warning light might appear on the dash – “check your tyres”.

That’s the concept.  The machine does things humans can do, only a lot faster, with a lot more data.

What does this mean for us drivers?  The machine helps be more accurate in our decisions, about when we refuel, and when the tyres need air.

There are many more complex and valuable uses for machine learning.  We’re working on one.

Our research targets a way of using that pattern recognition to guide clinicians in ways they care for patients.  We aren’t clinicians and this isn’t diagnosis – that’s above our pay grade.  But we are able to giveintensivists, surgeons and nurses a heads up.  When our machine spotted this particular set of readings in the past, 75% of the time thepatient suffered an episode within 30 minutes.

Our machine has learned when A+B/C =X a particular event followed within a short period of time.

original post at avantrasara.com