The short answer is, you can't. Solomonoff induction is not computable, and moreover depends on your model of computation. Asymptotically, the model of computation makes no difference, but for any finite set of examples, there is (for example) a universal Turing machine with short codes for those examples built in.
Practical methods of choosing among models use completely different methods. One collection of such methods is called regularization.
In practice no one honestly computes the complexity of models in any of the fields I am familiar with, such as physics, bio, chem etc. Sometimes they count the number of parameters/degrees of freedom, like in your example. In reality there is a dearth of models that explain observations and predict something new, interesting and testable, so the issue rarely arises.
Solomonoff induction is generally given as the correct way to penalise more complex hypotheses when calculating priors. A great introduction can be found here.
My question is, how is this actually calculated in practice?
As an example, say I have 2 hypotheses:
A. The probability distribution of the output is given by the same normal distribution for all inputs, with mean μ and standard deviation σ.
B. The probability distribution of the output is given by a normal distribution depending on an input x with mean μ0+mx and standard deviation σ.
It is clear that hypothesis B is more complex (using an additional input [x], having an additional parameter [m] and requiring 2 additional operations to calculate) but how does one calculate the actual penalty that B should be given vs A?