Note this is going to be the first of a rambling set of posts about my exploration into AI, I am going to write this as I find it, and at the end (or sensible points) will produce a summary of the bits you actually need.
While chatting to a fried of mine. we got around to the topic of Artificial Intelligence as one of the teams he worked with was looking into how it could be used for architecture design. This struck me as an interesting topic to investigate as AI is not something I have ever dabbled in. so where to start?
A while back I had a phase of buying books on humble bundles that seemed interesting, but most of which I never got round to reading, so lets take a look and see if the famous meme is really the case.
Deep Learning with TensorFlow – Second Edition
One that struck me at first was “Deep Learning with TensorFlow – Second Edition” I don’t know why that one caught my eye but it did, so lets see where this takes me.
Okay so we get into it and it’s based on ubuntu with a number of libraries, well I am on a PC nativly (yes I do a lot on the PI’s but I don’t think a Pi Zero is going to cut it). So off to run up a VM for Ubuntu, or my preference Linux Mint.
Okay VMware Player is not starting up, that’s not a good sign. Never mind I did see the Version 16 was out, so lets upgrade first. – That seemed to fix it.
Once that’s up and running we can run the PIP command to install the tools etc:
sudo pip3 install pandas numpy tensorflow sklearn seaborn tffm
Okay that didn’t work as the Linux mint doesn’t know what pip3 is, trying again:
sudo apt install python3-pip
sudo pip3 install pandas numpy tensorflow sklearn seaborn tffm
Close, but not quite:
sudo apt install python3-pip
sudo apt install python3-testresources
sudo pip3 install pandas numpy tensorflow sklearn seaborn tffm
that’s better.
Onto chapter 1. This is talking about ML and DL for Machine Learning and Deep Learning, but no mention of AI.
Machine Learning – Look at historic data, improve and come up with predictions for future.
Looks like there will be a lot of definitions and terms, so I will try to highlight them like this for clarity.
Well that’s certainly close to the IF statement chain, with one exception. The If chain is what the programmer writes based on their knowledge of the problems in the past and the program works in the future. This seems to include the changing of the rules as time goes on. Hey it’s the learning part.
Training set – the information used to get the initial setup for the algorithm.
Validation set – information used to tune the algorithm
Test set – information used to check the algorithm is working
Not 100% sure I understand this yet, so might come back and refine these.
Deep Learning – Seems to be about abstracting information to a higher level to allow it to handle bigger sets of information.
The next pages get more and more techno speak, and goodness me this is now getting very much into something that sounds almost like sentences but I am not sure I understand the language used.
Got to the end of chapter 1 and I think I need a more introductory book to make more sense of this. So lets take a look at “Artificial Intelligence By Example”.
Artificial Intelligence By Example
Okay this seems a lot less jargon and a lot more hands on, but let’s see where it goes.
First example Python code does run and produces an output that makes sense (to me anyway). Interestingly the results are identical to the ones in the book. Normally that is expected, but with there being a lot of random choices I would have expected there to be some variation to the end results.
Okay that’s the end of chapter 1 of the 2nd book. This seems a different approach and more into my existing understanding, so lets keep going. But that’s for another day.
Pauls simple guides 