giovgiac / TensorNewtonRaphson

An implementation of the numerical method for root finding of Newton-Raphson using TensorFlow for automatic derivative calculation and possibility for CUDA acceleration.

TensorNewtonRaphson (TNR)

This package implements the numerical method for root finding, Newton-Raphson, using the TensorFlow library.

Introduction

The Newton-Raphson is a method for numerically finding the roots of any given equation, provided the derivative exists, is known and the method succeeds in converging to the answer.

By making use of TensorFlow, this code allows the method to be run without having to calculate the derivative and also make use of GPU power through CUDA if necessary.

Testing

To verify if everything is functioning properly before making your own calculations, run the unittest of the module by executing the following in your command shell:

$ python test.py

Executing

To run the algorithm, all you have to do is configure the desired equation, initial guess and precision on the main.py file and then execute the following in your command shell:

$ python main.py

Or, if you prefer configuring it in the command shell, you can also execute the algorithm like this:

$ python main.py --initial_guess=1.0 --precision=1e-9 --equation="tf.pow(x, 2)"