neurosyntax / policy-smoothing Goto Github PK

Implement the pricing information provided by AWS as a Python function that takes resource quantities as an input and produces the total cost (or costs) as an output.

Leave ideas, links and etc. here!

** not urgent: for down the road

• Train a classifier on a bunch of different functions and how well a range of neural network topologies approximate that function

  • The input to this classifier would be some representation of the function's AST, and the expected output would be which class of neural network should be used to approximate the function. Potentially more powerful than trying to define families of functions/ASTs explicitly.

• We can just generate a whole bunch of small ASTs to use as training data for this classifier

Given what we know about the pricing algorithm for the cloud use case, determine what would be a suitable class of neural network and how it can be used/implemented.

Define Python decorators that take domain information as parameters and can be applied to a Python function definition in order to induce automated training of a neural network on that function (where training would occur in the application itself).

Find a suitable representation for a neural networks and, upon completing of training for a given function, cache the neural network (e.g., on disk) such that it can be reused if no changes were made between different script runs.

For an assortment of simple functions, including:

• Max/min
• Averages
• And/Or
• Vector norms

Would be valuable to see how different kinds of networks perform at approximating different kinds of functions.

Just to make everything more consistent as we start to push out more code, could we settle on some style conventions? Some things come to mind:

• 4 spaces for indentation
• 80 character line limit
• camel case for classes (e.g., MyClass), snake case for everything else (e.g., my_variable, my_function) (and I suppose dashes for file/directory names, like policy-smoothing).
• docstring for every function explaining at a minimum what it does, and preferably what inputs are expected, and what it returns and possibly what exceptions it raises.

Are these ok with everyone, and does anyone have any to add? We could/should eventually put these into a readme.

Also, we should also settle on what versions of things we're using. In particular, @benlawson 's amazon pricing script uses Python 3, but @ch3njust1n 's code uses Python 2. Is there a reason not to use Python 3 for everything?