Deterministic subsets should use Herman's ordered subset method

[robbietuk]

This method is better for reconstruction performance.

[1] G. T. Herman and L. B. Meyer, “Algebraic Reconstruction Techniques Can Be Made Computationally Efficient” IEEE Trans. Med. Imaging, vol. 12, no. 3, pp. 600–609, 1993, doi: 10.1109/42.241889.

Hackathon-000-Stochastic-Algorithms/cil/NewSubsetSumFunction.py

Lines 168 to 172 in 72f468c

	if self.deterministic == True:
	if self.subset_num != self.num_subsets:
	self.subset_num += 1
	else:
	self.subset_num = 1

[robbietuk]

Acutually, this method shouldn't use the determinisitc bool at all and instead a subset_select_function should be passed to the class

Hackathon-000-Stochastic-Algorithms/cil/NewSubsetSumFunction.py

Lines 119 to 131 in 72f468c

	class SubsetSumFunction(AveragedSumFunction):
	'''Class for use as objective function in gradient type algorithms to enable the use of subsets.
	The `gradient` method is implemented in children classes and allows to return an approximation of the gradient based on subset gradients.
	Parameters:
	-----------
	- functions: F_{1}, F_{2}, ..., F_{num_subsets}
	- (optional)
	subset_select_function: function which takes two integers and outputs an integer
	defines how the subset is chosen at each iteration
	default is uniformly at random
	'''
	def __init__(self, functions, subset_select_function=(lambda a: int(np.random.choice(a))), replacement = False, deterministic = False, subset_init=-1, **kwargs):

This subset_select_function should perform the deterministic subset selection methodology.