many bug fixes and improvemnents to xfit

+xolotl/+examples/burstingCostFcn.m

@@ -26,9 +26,9 @@
 
 
 % accumulate errors
-C = xfit.binCost([950 1050],metrics.burst_period);
-C = C + xfit.binCost([.1 .3],metrics.duty_cycle_mean);
-C = C + xfit.binCost([7 10],metrics.n_spikes_per_burst_mean);
+C = xtools.binCost([950 1050],metrics.burst_period);
+C = C + xtools.binCost([.1 .3],metrics.duty_cycle_mean);
+C = C + xtools.binCost([7 10],metrics.n_spikes_per_burst_mean);
 
 % safety -- if something goes wrong, return a large cost
 if isnan(C)

+xolotl/testConductances.m

@@ -8,6 +8,7 @@
 if exist([xroot filesep 'cond_check.mat'],'file') == 2
 	load([xroot filesep 'cond_check.mat'])
 else
+	warning('It looks like this is the first time conductances are being checked. This will take a while...')
 	ok_channels = {};
 end
 

+xtools/binCost.m

@@ -0,0 +1,46 @@
+%        __ _ _
+% __  __/ _(_) |_
+% \ \/ / |_| | __|
+%  >  <|  _| | |_
+% /_/\_\_| |_|\__|
+%
+%
+% ### binCost
+%
+% **Syntax**
+%
+% ```matlab
+% c = xfit.binCost(allowed_range, actual_value)
+% ```
+%
+% **Description**
+%
+% A static method that computes a quadratic cost
+% when `actual_value` is outside of the minimum and maximum
+% set by the 2-vector `allowed_range`, and returns zero otherwise.
+% 
+% This method can be used as part of a simulation function
+% when it is important for neurocomputational properties of interest to
+% fit within a given range, rather than necessarily fit a value perfectly.
+
+function c = binCost(allowed_range,actual_value)
+
+
+if isnan(actual_value)
+	c = 1;
+	return
+end
+
+w = (allowed_range(2) - allowed_range(1))/2;
+m = (allowed_range(2) + allowed_range(1))/2;
+
+if actual_value < allowed_range(1)
+	d = m - actual_value;
+	c = (1- (w/d));
+elseif actual_value > allowed_range(2)
+	d = actual_value - m;
+	c = (1- (w/d));
+else
+	% no cost
+	c = 0;
+end

@xfit/evaluate.m

@@ -24,7 +24,7 @@ and return a cost (a double).
 configured in the xfit object:
 
 - `x` (the xolotl object)
-- `sim_func`
+- `SimFcn`
 - `parameter_names`
 
 %}
@@ -34,4 +34,4 @@ and return a cost (a double).
 for i = 1:length(self.parameter_names)
 	self.x.set(self.parameter_names{i},params(i))
 end
-c = self.sim_func(self.x, self.data);
+c = self.SimFcn(self.x, self.data);

@xfit/fit.m

@@ -29,7 +29,7 @@ function fit(self)
 
 assert(~isempty(self.parameter_names),'No parameter names defined')
 assert(~isempty(self.x),'Xolotl object not configured')
-assert(~isempty(self.sim_func),'Simulation function not set')
+assert(~isempty(self.SimFcn),'Simulation function not set')
 
 if isempty(self.seed) && ~isempty(self.ub) && ~isempty(self.lb)
 	% pick a random seed within bounds

@xfit/xfit.m

@@ -10,33 +10,38 @@
 	x@xolotl
 
 	% function to minimize
-	sim_func@function_handle
+	SimFcn@function_handle
 
 	% parameters to optimize
 	parameter_names@cell
 	seed
 	lb
 	ub
 
+	PlotFcn
+
 
 
-	display_type = 'iter'
+	Display = 'iter'
 	engine 
 
-	% logging
-	timestamp
 
 
 	% this can be used to store any user-defined data
 	data
 
 
+	options
+
 end % end props
 
 
 properties (SetAccess = private)
 	best_cost
-	options
+
+
+	% logging
+	timestamp
 end
 
 methods

examples/demo_xfit.m

@@ -9,9 +9,11 @@
 % we instantiate the xfit object
 % and specify the particle swarm engine
 p = xfit('particleswarm');
+p.options.UseParallel = true;
 
 % we assign a cost function
-p.sim_func = @xolotl.examples.burstingCostFcn;
+p.SimFcn = @xolotl.examples.burstingCostFcn;
+
 
 % we assign the xolotl object
 p.x = x;
@@ -34,6 +36,10 @@
 plot(time,V,'k')
 title('Before optimization')
 
+subplot(2,1,2); hold on
+set(gca,'XLim',[0 10],'YLim',[-80 50])
+
+
 % perform the optimization procedure
 p.fit;
 
@@ -43,8 +49,7 @@
 
 
 % visualize the results of the optimization
-subplot(2,1,2); hold on
-set(gca,'XLim',[0 10],'YLim',[-80 50])
+
 x.t_end = 10e3;
 V = x.integrate;
 time = (1:length(V))*1e-3*x.dt;