1 %> @brief Shows different ways to paint the same Feature Histogram
23 % The
object that will
do the feature selection
25 ofs.nf_select = 10; % <-------- Number of features to be selected
26 ofs.fsg = fsg_clssr01;
31 osgs.flag_perclass = 0;
33 osgs.no_reps = 50; % <-------- Number of repetitions
for the histogram
35 % The Feature Selection Repeater
42 log_rep = orep.use(ds); % This is the time-consuming line
46 %> Stability and nf x grade
50 ds_nfxgrade = log_rep.extract_dataset_nfxgrade();
51 ds_stab = log_rep.extract_dataset_stabilities();
72 % Plots histogram in 3 different styles
76 %> All features are informative
78 ssp.nf4gradesmode =
'fixed';
79 ov.colors = []; % Uses
default color scheme from
colors2map(), which accesses COLOR_STACKEDHISTS
81 %> 4 features are informative
83 ssp.nf4gradesmode =
'fixed';
84 ssp.stabilitythreshold = 0.05;
87 %> Same
as previous but with different color scheme
89 ssp.nf4gradesmode =
'fixed';
90 ssp.stabilitythreshold = 0.05;
91 ov.colors = {[.6, 0, 0], [1, 0, 0], [.7, .7, .7], [.9, .9, .9]};
93 log_ssp = ssp.use(log_rep);
96 ov.use(log_ssp); % .draw_stackedhists(ds, colors, []);
function data_select_hierarchy(in data, in hierarchy)
Linear and Quadratic discriminant.
function maximize_window(in h, in aspectratio, in normalizedsize)
Visualization - All curves in dataset.
function colors2map(in colors, in no_hists, in no_informative)
Visualization - Stacked histograms.
function save_as_png(in h, in fn, in dpi)
Forward Feature Selection.
Analysis Session - Feature Selection Repeater.
FSG that uses classifier to grade the subsets.
function load_data_she5trays()
Processor of a set of subsets of features.
Analysis Session (AS) base class.