rs_xform_coordsets_demo

Demonstrate specification and application of simple linear transformations

plot options illustrated: choice of dimension and coordinates to plot several plots into same figure custom arrangement of subplots custom colors for data points custom marker sizes for points custom symbol choice custom symbol fills custom alpha blending custom axis labels custom axis ranges custom labeling of datasets based on subject ID and paradigm name callouts for stimulus label selection of data points to label based on length of stimulus name custom choice of data points to connect between sets custom line style and color for connections between sets rotation of raw data coordinates into a consensus

Also illustrates: silencing logging for rs_[get|align|knit]_coordsets, saving intermediate results from alignment so that alignment is not repeated by rs_knit_coordsets

See also: rs_xform_specify, rs_xform_apply, rs_disp_coorddsets

section to force btc defaults, even if rs_aux_defaults.mat has been created or modified

if ~exist('aux_force_filename') aux_force_filename='rs_aux_defaults_btc.mat'; end
auxs_force=struct;
opts_needed={'opts_read','opts_rays','opts_check','opts_align','opts_qpred','opts_knit','opts_xform','opts_disp'};
for k=1:length(opts_needed)
    auxs_force.(opts_needed{k})=rs_aux_force(opts_needed{k},[],aux_force_filename);
end

filenames={'./samples/animals/image_coords_S5','./samples/animals/word_coords_S5'};
nsets=length(filenames);
aux_in=auxs_force;
aux_in.opts_read=setfields(aux_in.opts_read,{'input_type','if_auto','if_log'},{1,1,0});
aux_in.nsets=nsets;

raw dataset

[data_read,aux_read]=rs_get_coordsets(filenames,aux_in);

Output:

##### rs_warning: cannot find stimulus coordinates, so cannot identify rays
##### rs_warning: cannot find stimulus coordinates, so cannot identify rays

align data

aux_align_def=auxs_force.opts_align;
aux_align_def.opts_align.if_log=0; %turn off logging
[data_align,aux_align]=rs_align_coordsets(data_read,aux_align_def);

Output:

##### rs_warning: cannot find stimulus coordinates, so cannot identify rays
##### rs_warning: cannot find stimulus coordinates, so cannot identify rays

keep intermediate results from alignment so that alignment isn't redone

aux_knit_def=auxs_force.opts_knit;
aux_knit_def.data_align=data_align;
aux_knit_def.sa_pooled=aux_align.sa_pooled;
aux_knit_def.opts_knit.if_log=0; %turn off logging
[data_consensus,aux_knit]=rs_knit_coordsets(data_align,aux_knit_def);
aux_knit_pc=aux_knit_def;
aux_knit_pc.opts_knit.if_pca=1; %turn on pca rotation

Output:

##### rs_warning: cannot find stimulus coordinates, so cannot identify rays

rotate to consensus and pc rotation

[data_consensus_pc,aux_knit_pc]=rs_knit_coordsets(data_align,aux_knit_pc);

Output:

##### rs_warning: cannot find stimulus coordinates, so cannot identify rays

set_labels=cell(1,2*nsets); %plot two original datasets and two transformed datasets
for iset=1:nsets
    set_labels{iset}=cat(2,data_read.sets{iset}.subj_id,': ',data_read.sets{iset}.paradigm_name);
    set_labels{nsets+iset}=cat(2,'transformed ',set_labels{iset});
end

label the stimuli with short names

label_maxlength=5; %max length of a stimulus label
data_label_list=[];
nstims=data_align.sas{1}.nstims;
for istim=1:nstims
    if length(data_align.sas{1}.typenames{istim})<=label_maxlength
        data_label_list(end+1)=istim;
    end
end

dlist=[2 3]; %dimensions to show
nds=length(dlist);
data_start={'raw','raw','raw','consensus','consensus','consensus','consensus','consensus_pc'};
nxforms=length(data_start);
aux_outs=cell(2,nds,nxforms); %d1: init or transformed, d2: dimension, d3: which transform

xform=cell(1,nxforms);
aux_spec_outs=cell(1,nxforms);
aux_xform_outs=cell(1,nxforms);
data_xform=cell(1,nxforms);

for ixform=1:nxforms
    switch data_start{ixform}
        case 'raw'
            data_use=data_align; %same as data_read, but stimuli in same order as consensus
            axis_label_prefix='dim';
        case 'consensus'
            data_use=aux_knit.components;
            axis_label_prefix='dim (consensus)';
        case 'consensus_pc'
            data_use=aux_knit_pc.components;
            axis_label_prefix='pc (consensus)';
    end
    opts_xform=auxs_force.opts_xform;
    switch ixform
        case 1 %specify linear transformations by hand for dims 2 and 3
            opts_xform.mode=[];
            xform{ixform}.pipeline=[];
            ts=cell(1,3);
            ts{2}.b=0.9;
            ts{2}.T=[cos(0.1) sin(0.1);-sin(0.1) cos(0.1)];
            ts{2}.c=[-.2 -.3];
            ts{3}.b=0.7;
            ts{3}.T=[cos(0.1) sin(0.1) 0;-sin(0.1) cos(0.1) 0;0 0 1.2]; %rotate on coords 1 and 2, magnify on coord 3
            ts{3}.c=[-.2 -.3 .5];
            xform{ixform}.ts{1}=ts;
            desc='rotate in dims 1 and 2, dliate dim 3';
        case 2
            opts_xform.mode='translate';
            opts_xform.source='global';
            opts_xform.centering_specifier='value';
            opts_xform.centering_value=0.5*[1:10];
        case 3
            opts_xform.mode='translate';
            opts_xform.source='local';
            opts_xform.centering_specifier='typename';
            opts_xform.centering_typename='ant';
        case 4
            opts_xform.mode='offset_pca';
            opts_xform.source='local';
            opts_xform.centering_specifier='none';
            opts_xform.centering_typename='snake'; %just for labeling, ignored in specification of transformation
        case 5
            opts_xform.mode='offset_pca';
            opts_xform.source='local';
            opts_xform.centering_specifier='typename';
            opts_xform.centering_typename='snake';
        case 6
            opts_xform.mode='offset_pca';
            opts_xform.source='global';
            opts_xform.centering_specifier='typename';
            opts_xform.centering_typename='snake';
        case {7,8}
            opts_xform.mode='translate_then_pca';
            opts_xform.source='global';
            opts_xform.centering_specifier='typename';
            opts_xform.centering_typename='snake';
    end

specify the transformation

    xform_name=sprintf('transformation %1.0f, starting with %s',ixform,data_start{ixform});
    if isempty(xform{ixform})
        [xform{ixform},aux_spec_outs{ixform}]=rs_xform_specify(data_use,setfield(auxs_force.opts_xform,'opts_xform',opts_xform));
        desc=sprintf('mode: %s (%s), centering specified by %s',opts_xform.mode,opts_xform.source,opts_xform.centering_specifier);
        if isfield(opts_xform,'centering_typename')
            desc=cat(2,desc,sprintf(' (%s)',opts_xform.centering_typename));
        end
    end

do the transformations

    [data_xform{ixform},aux_xform_outs{ixform}]=rs_xform_apply(data_use,xform{ixform},setfield(struct(),'opts_xform',auxs_force.opts_xform));

plot: top row is untransformed data, dims 2 and 3; bottom row is transfornmed data

    hfig=figure;
    haxes=cell(1,nds);
    for id=1:nds
        haxes{1,id}=subplot(1,nds,id);
    end
    opts_disp_init=auxs_force.opts_disp;
    opts_disp_init.fig_handle=hfig;
    opts_disp_init.fig_name=xform_name;
    opts_disp_init.axis_label_prefix=axis_label_prefix;
    opts_disp_init.set_labels=set_labels;
    opts_disp_init.set_colors={'r',[0 0.6 0.1]'};
    opts_disp_init.set_markers={'o'};
    opts_disp_init.set_filled=[1 1 0 0]; %transformed sets are unfilled
    opts_disp_init.set_alphas=[0.3 0.3 1 1]; %untransformed sets are alpha-blended
    opts_disp_init.set_markersizes=6;
    opts_disp_init.data_label_method='list';
    if strcmp(opts_xform.mode,'offset_pca') | strcmp(opts_xform.mode,'translate_then_pca') %just label the stimlus used for offset, and one more stimulus
        opts_disp_init.data_label_list=[1,strmatch(opts_xform.centering_typename,data_align.sas{1}.typenames,'exact')];
    else
        opts_disp_init.data_label_list=data_label_list;
    end
    opts_disp_init.data_label_font_size=7;
    opts_disp_init.callout_amount=0.2;
    opts_disp_init.connect_sets_data_method='labeled';
    opts_disp_init.connect_sets_linestyles={':'};
    opts_disp_init.connect_sets_color_mode='list';
    opts_disp_init.connect_sets_colors='k';

    if strcmp(data_start{ixform},'raw')
        opts_disp_init.data_label_setsel_method='all';
        opts_disp_init.connect_sets_method='list';
        opts_disp_init.connect_sets_list=[1 3;2 4]; %connect each dataset with its transform
        opts_disp_init.axis_range='tight';
    else
        opts_disp_init.data_label_setsel_method='list';
        opts_disp_init.data_label_setsel_list=[1 2];
        opts_disp_init.connect_sets_method='list';
        opts_disp_init.connect_sets_list=[1 2;1 3;2 4]; %connect the two raw datasets to eachother and each one to its transform
        opts_disp_init.axis_range='list';
        opts_disp_init.axis_range_list=[-10 10;-7 7;-5 5]; %fixed scales to make transforms easier to see
    end
    data_disp=auxs_force.opts_disp;
    data_disp.ds=[data_use.ds,data_xform{ixform}.ds];
    data_disp.sas=[data_use.sas,data_xform{ixform}.sas];
    data_disp.sets=[data_use.sets,data_xform{ixform}.sets];
    for id=1:nds
        opts_disp_init.axis_handles=haxes(1,id);
        opts_disp_init.dim_select=dlist(id);
        aux_outs{1,id,ixform}=rs_disp_coordsets(data_disp,setfield(struct,'opts_disp',opts_disp_init));
    end
    axes('Position',[0.01,0.03,0.01,0.01]);
    text(0,0,desc,'Interpreter','none');
    axis off
    axes('Position',[0.01,0.06,0.01,0.01]);
    text(0,0,xform_name,'Interpreter','none');
    axis off
end %ixform