Installing desolver

desolver is a pure python module that builds upon the functionalities provided by numpy for a relatively efficient numerical integration library.

Furthermore, desolver incoporates the use of pyaudi and pytorch for more advanced applications.

With these libraries, it is possible to incorporate gradient descent of parameters into the numerical integration of some system. Differential Intelligence example shows how to use pyaudi to tune the weights of a controller based on the outcome of a numerical integration.

To install desolver simply type:

pip install desolver

Refer to the following links for pyaudi support and pytorch support. desolver will automatically detect these modules if they are in the same environment and use them if possible.

It is necessary to tell desolver if you’d like to use numpy (and pyaudi if available) or pytorch as the backend for the computations. To do this, you can set the environment variable DES_BACKEND to either numpy or torch as shown in the snippet below:

import os
os.environ['DES_BACKEND'] = 'torch' # or 'numpy'

import desolver

To check that all went well fire-up your python console and try the example in quick-start example.