A symbolic algebra library for multivariate Gaussian and Gaussian process probabilistic models.

See project report in documents.

Install from PyPi using the following command:

sudo pip install SymGP

See examples in folder notebooks.

• Allow utils.replace to replace expressions of the form D+A in D+2*A with C to give A+C.

• Provide functionality for utils.simplify to automatically replace expressions composed of KernelMatrix matrices e.g.
  for an expression K(f,u)*K(u,u).I*K(u,g), we would want to replace the whole expression with a new Kernel
Q = Kernel(sub_kernels=[K,K],kernel_type='mul',mat=M,name='Q') (see kernel/kernel.py) where
  M = Constant('M_{u,u}',n,n,full_expr=K(u,u).I), n = Symbol('n') (represents a variable shape)
  K = Kernel() (represents a generic kernel function). This would then give the expression Q(f,g)

  This can be done using utils.replace by calling utils.replace(K(f,u)*K(u,u).I*K(u,g),[Q]) but we need to find
  a way to detect these types of expressions automatically in simplify.

• Add more functionality to the MVG class to allow for example, addition of MVG objects.

• Improve code generation to convert symbolic expressions to code files in languages such as Python, Matlab, Julia, etc.