askadam.optmisation

Usage

obj = askadam;
out = obj.optimisation( data, mask, weights, parameters, fitting, FWDfunc, varargin);

I/O overview

Input	Description
data	(Masked) N-D (imaging) data
mask	(1-3)D signal mask applied on FWDfunc, NOTE this mask does NOT apply on data
weights	N-D wieghts, same dimension as ‘data’ (optional)
parameters	structure variable containing starting points of all model parameters to be estimated (optional)
fitting	structure contains fitting algorithm parameters
fitting.optimiser	Algorithm for parameter update, ‘adam’ (default) \| ‘sgdm’ \| ‘rmsprop’
fitting.model_params	1xM cell variable, name of the model parameters, e.g. {‘S0’,’R2star’};
fitting.lb	1xM numeric variable, fitting lower bound, same order as field ‘model_params’, e.g. [0.5, 0];
fitting.ub	1xM numeric variable, fitting upper bound, same order as field ‘model_params’, e.g. [2, 1];
fitting.isdisplay	boolean, display optimisation process in graphic plot
fitting.convergenceValue	tolerance in loss gradient to stop the optimisation, set to -inf to disable checking the gradient; +ve: decreasing loss; -ve: increasing loss
fitting.convergenceWindow	#iterations in which ‘convergenceValue’ is computed
fitting.patience	#iterations to wait before stopping then optimisation when the loss does not improve
fitting.iteration	maximum # of optimisation iterations
fitting.initialLearnRate	(initial) learn rate of Adam optimiser
fitting.tol	tolerance in loss
fitting.lambda	regularisation parameter(s)
fitting.regmap	model parameter(s) in which regularisation is applied
fitting.TVmode	Mode for total variation (TV) regularisation, ‘2D’ \| ‘3D’
fitting.lossFunction	loss function, ‘L1’ \| ‘L2’ \| ‘huber’ \| ‘mse’
fitting.randomness	randomness of starting point; value between [0,1] 0 (fixed point) > somewhere inbetween >1 completely random
fitting.debug	Dispay extra messages for debugging
FWDfunc	function handle for forward signal generation; size of the output must match size of ‘data’
varargin	additional input for FWDfunc other than ‘parameter’ and ‘mask’ (same order as FWDfunc)

Output	Description
out	structure contains optimisation result
out.final	output structure at final iteration
out.final.loss	total loss = loss_fidelity + loss_reg
out.final.loss_fidelity	loss of data consistency term
out.final.loss_reg	loss of regularisation term
out.final.(model_params{k})	estimated model parameter(s)
out.min	output structure at loss is minimum