如何在 Matlab 中更正此错误
How to correct this error in Matlab
我在下面自定义了部分函数 feature_select.m，以使用我自己生成的掩码 whole_brain_mask_1、whole_brain_mask_2、…、whole_brain_mask_10 到 subj。
function [subj] = feature_select(subj,data_patin,regsname,selsgroup,varargin)
% No-peeking feature selection
%
% [SUBJ] = FEATURE_SELECT(SUBJ,DATA_PATIN,REGSNAME,SELSGROUP,...)
%
% Calls a statmap generation function multiple times, using
% a different selector each time. This creates a group of
% statmaps, which are then thresholded to create a group of
% boolean masks, ready for use in no-peeking
% cross-validation classification.
%
% Adds the following objects:
% - pattern group of statmaps called NEW_MAP_PATNAME
% - mask group based on the statmaps called
% sprintf('%s%i',NEW_MASKSTEM,THRESH)
%
% DATA_PATIN should be the name of the pattern object that
% contains voxel (or other feature) values that you want to
% create a mask of. If DATA_PATIN is a group_name, then this
% will use a different member of the group for each
% iteration.
%
% REGSNAME should be a binary nConds x nTimepoints 1-of-n matrix
%
% SELSGROUP should be the name of a selectors group, such as
% created by create_xvalid_indices
%
% For each iteration: call the ANOVA on the DATA_PATIN data,
% which will produce a statmap, employing only the TRs
% labelled with a 1 in the selector for that iteration
%
% NEW_MAP_PATNAME (optional, default = DATA_PATIN +
% STRIPPED_NAME). The name of the new statmap pattern group
% to be created. By default, this will be 'anova' if
% STATMAP_FUNCT = 'statmap_anova' etc.
%
% NEW_MASKSTEM (optional, default = DATA_PATIN +
% 'anovathresh'). The name of the new thresholded boolean
% mask group to be created from the ANOVA statmap. You'll
% need to create multiple mask groups if you want to try out
% multiple thresholds, so adding the threshold to the name
% is a good idea
%
% THRESH (optional, default = 0.05). Voxels that don't meet
% this criterion value don't get included in the boolean
% mask that gets created from the ANOVA statmap. If THRESH =
% [], the thresholding doesn't get run
%
% STATMAP_FUNCT (optional, default = 'statmap_anova'). Feed
% in a function name and this will create a function handle
% to that and use it to create the statmaps instead of
% statmap_anova
%
% STATMAP_ARG (optional, default = []). If you're using an
% alternative voxel selection method, you can feed it a
% single argument through this
%
% Need to implement a THRESH_TYPE argument (for p vs F
% values), which would also set the toggle differently xxx
%
% e.g. subj = feature_select( ...
% subj,'epi_z','conds','runs_nmo_xvalid','thresh',0.001)
% License:
%=====================================================================
%
% This is part of the Princeton MVPA toolbox, released under
% the GPL. See http://www.csbmb.princeton.edu/mvpa for more
% information.
%
% The Princeton MVPA toolbox is available free and
% unsupported to those who might find it useful. We do not
% take any responsibility whatsoever for any problems that
% you have related to the use of the MVPA toolbox.
%
% ======================================================================
defaults.new_map_patname = '';
defaults.new_maskstem = sprintf('%s_thresh',data_patin);
defaults.thresh = 0.05;
defaults.statmap_funct = 'statmap_anova';
defaults.statmap_arg = struct([]);
args = propval(varargin,defaults);
if isempty(args.new_map_patname)
% get the name of the function being run, e.g. 'statmap_anova' -> 'anova'
stripped_name = strrep(args.statmap_funct,'statmap_','');
args.new_map_patname = sprintf('%s_%s',data_patin,stripped_name);
end
% append the thresh to the end of the name
args.new_maskstem = sprintf( ...
'%s%s',args.new_maskstem,num2str(args.thresh));
% Find the selectors within the specified group
selnames = find_group(subj,'selector',selsgroup);
nIterations = length(selnames);
[data_patnames isgroup] = find_group_single(subj,'pattern',data_patin,'repmat_times',nIterations);
if length(data_patnames) ~= length(selnames)
error('Different number of patterns and selectors');
end
if nIterations == 0
error('No selectors in %s group',selsgroup);
end
% % this warning used to be here to remind people of the
% % existence of peek_feature_select, but since there are good
% % reasons why one might want to have just one selector
% % without using peek_feature_select, i took it out
% if nIterations == 1
% warning('You''re only calling the anova once because you have one selector - use peek_feature_select instead?');
% end
if ~ischar(args.statmap_funct)
error('The statmap function name has to be a string');
end
disp( sprintf('Starting %i %s iterations',nIterations,args.statmap_funct) );
for n=1:nIterations
fprintf(' %i',n);
% Get the pattern for this iteration
cur_data_patname = data_patnames{n};
% Get the selector name for this iteration
cur_selname = selnames{n};
% Name the new statmap pattern and thresholded mask that will be created
cur_maskname = sprintf('%s_%i',args.new_maskstem,n);
cur_map_patname = sprintf('%s_%i',args.new_map_patname,n);
% if a pattern with the same name already exists, it
% will trigger an error later in init_object, but we
% want to catch it here to save running the entire
% statmap first
if exist_object(subj,'pattern',cur_map_patname)
error('A pattern called %s already exists',cur_map_patname);
end
if ~isempty(args.statmap_arg) && ~isstruct(args.statmap_arg)
warning('Statmap_arg is supposed to be a struct');
end
% Add the current iteration number to the extra_arg, just in case
% it's useful
args.statmap_arg(1).cur_iteration = n;
% Create a handle for the statmap function handle and then run it
% to generate the statmaps
statmap_fh = str2func(args.statmap_funct);
subj = statmap_fh(subj,cur_data_patname,regsname,cur_selname,cur_map_patname,args.statmap_arg);
subj = set_objfield(subj,'pattern',cur_map_patname,'group_name',args.new_map_patname);
if ~isempty(args.thresh)
% Now, create a new thresholded binary mask from the p-values
% statmap pattern returned by the anova
subj = create_thresh_mask(subj,cur_map_patname,cur_maskname,args.thresh);
subj = set_objfield(subj,'mask',cur_maskname,'group_name',args.new_maskstem);
end
end % i nIterations
disp(' ');
disp( sprintf('Pattern statmap group ''%s'' and mask group ''%s'' created by feature_select', ...
args.new_map_patname,args.new_maskstem) );
这是上面函数的一部分，我已将其改编为使用我自己的掩码：
dirIn_Mask = ['/D disk/my_fold/My_generated_Mask'];
data_patin = subj.patterns{1,2}.name;
defaults.new_maskstem = sprintf(‘whole_brain_mask’);
defaults.statmap_arg = struct([]);
args = propval(varargin,defaults);
args.new_maskstem = sprintf( '%s',args.new_maskstem );
for n = 1:1:10
fprintf(' %i', n );
inputFile_mask = ['whole_brain_mask_',num2str(n)];
load(fullfile(dirIn_Mask,inputFile_mask));
cur_maskname = sprintf('%s_%i',args.new_maskstem,n );
subj.masks{1,(runID+1)}.mat = whole_brain_new_1;
subj = set_objfield(subj,'mask',cur_maskname,'group_name',new_maskstem);
end
这是我得到的错误：
Attempt to execute SCRIPT varargin as a function:
C:\D disk\MATLAB\R2014a\toolbox\matlab\lang\varargin.m
Error in MVPA (line 89)
args = propval(varargin,defaults);
谁能帮我解决这个问题？
从您的代码中删除 varargin。 varargin 是一个特殊关键字，仅当您 运行 将代码作为函数时才可用。 varargin 允许将可变数量的输入放入您的函数中。这些输入被散列到一个元胞数组中。
因为这是一个脚本文件，所以那个变量不存在。从您收到的 MATLAB 错误中也可以看出这一点。 FWIW，MATLAB 错误消息非常冗长。很难误解或误解他们想说的话。
因此，这样做：
args = propval(defaults);
如果这不起作用，那么我不确定 propval 是如何接受输入的，但也许可以通过放入一个空元胞数组来愚弄它：
args = propval(cell(1,1), defaults);
因为函数 propval 依赖于可变数量的输入，它应该处理 varargin 不存在的情况，所以我们可以模拟 varargin 没有额外的使用空元胞数组的参数。
您正在使用 varargin 作为 propval 的参数，但 varargin 未在您的代码中的任何位置定义。在这种情况下，varargin 回滚到 Matlab 脚本 varargin。结果，解释器假设你想 运行 它作为一个函数，returns 值到 propval 并且假设脚本不应该 return 值，你得到错误信息。
附带说明：varargin 通常设置为函数的最后一个参数，以允许程序员声明具有可变数量参数的函数。
我在下面自定义了部分函数 feature_select.m，以使用我自己生成的掩码 whole_brain_mask_1、whole_brain_mask_2、…、whole_brain_mask_10 到 subj。
function [subj] = feature_select(subj,data_patin,regsname,selsgroup,varargin)
% No-peeking feature selection
%
% [SUBJ] = FEATURE_SELECT(SUBJ,DATA_PATIN,REGSNAME,SELSGROUP,...)
%
% Calls a statmap generation function multiple times, using
% a different selector each time. This creates a group of
% statmaps, which are then thresholded to create a group of
% boolean masks, ready for use in no-peeking
% cross-validation classification.
%
% Adds the following objects:
% - pattern group of statmaps called NEW_MAP_PATNAME
% - mask group based on the statmaps called
% sprintf('%s%i',NEW_MASKSTEM,THRESH)
%
% DATA_PATIN should be the name of the pattern object that
% contains voxel (or other feature) values that you want to
% create a mask of. If DATA_PATIN is a group_name, then this
% will use a different member of the group for each
% iteration.
%
% REGSNAME should be a binary nConds x nTimepoints 1-of-n matrix
%
% SELSGROUP should be the name of a selectors group, such as
% created by create_xvalid_indices
%
% For each iteration: call the ANOVA on the DATA_PATIN data,
% which will produce a statmap, employing only the TRs
% labelled with a 1 in the selector for that iteration
%
% NEW_MAP_PATNAME (optional, default = DATA_PATIN +
% STRIPPED_NAME). The name of the new statmap pattern group
% to be created. By default, this will be 'anova' if
% STATMAP_FUNCT = 'statmap_anova' etc.
%
% NEW_MASKSTEM (optional, default = DATA_PATIN +
% 'anovathresh'). The name of the new thresholded boolean
% mask group to be created from the ANOVA statmap. You'll
% need to create multiple mask groups if you want to try out
% multiple thresholds, so adding the threshold to the name
% is a good idea
%
% THRESH (optional, default = 0.05). Voxels that don't meet
% this criterion value don't get included in the boolean
% mask that gets created from the ANOVA statmap. If THRESH =
% [], the thresholding doesn't get run
%
% STATMAP_FUNCT (optional, default = 'statmap_anova'). Feed
% in a function name and this will create a function handle
% to that and use it to create the statmaps instead of
% statmap_anova
%
% STATMAP_ARG (optional, default = []). If you're using an
% alternative voxel selection method, you can feed it a
% single argument through this
%
% Need to implement a THRESH_TYPE argument (for p vs F
% values), which would also set the toggle differently xxx
%
% e.g. subj = feature_select( ...
% subj,'epi_z','conds','runs_nmo_xvalid','thresh',0.001)
% License:
%=====================================================================
%
% This is part of the Princeton MVPA toolbox, released under
% the GPL. See http://www.csbmb.princeton.edu/mvpa for more
% information.
%
% The Princeton MVPA toolbox is available free and
% unsupported to those who might find it useful. We do not
% take any responsibility whatsoever for any problems that
% you have related to the use of the MVPA toolbox.
%
% ======================================================================
defaults.new_map_patname = '';
defaults.new_maskstem = sprintf('%s_thresh',data_patin);
defaults.thresh = 0.05;
defaults.statmap_funct = 'statmap_anova';
defaults.statmap_arg = struct([]);
args = propval(varargin,defaults);
if isempty(args.new_map_patname)
% get the name of the function being run, e.g. 'statmap_anova' -> 'anova'
stripped_name = strrep(args.statmap_funct,'statmap_','');
args.new_map_patname = sprintf('%s_%s',data_patin,stripped_name);
end
% append the thresh to the end of the name
args.new_maskstem = sprintf( ...
'%s%s',args.new_maskstem,num2str(args.thresh));
% Find the selectors within the specified group
selnames = find_group(subj,'selector',selsgroup);
nIterations = length(selnames);
[data_patnames isgroup] = find_group_single(subj,'pattern',data_patin,'repmat_times',nIterations);
if length(data_patnames) ~= length(selnames)
error('Different number of patterns and selectors');
end
if nIterations == 0
error('No selectors in %s group',selsgroup);
end
% % this warning used to be here to remind people of the
% % existence of peek_feature_select, but since there are good
% % reasons why one might want to have just one selector
% % without using peek_feature_select, i took it out
% if nIterations == 1
% warning('You''re only calling the anova once because you have one selector - use peek_feature_select instead?');
% end
if ~ischar(args.statmap_funct)
error('The statmap function name has to be a string');
end
disp( sprintf('Starting %i %s iterations',nIterations,args.statmap_funct) );
for n=1:nIterations
fprintf(' %i',n);
% Get the pattern for this iteration
cur_data_patname = data_patnames{n};
% Get the selector name for this iteration
cur_selname = selnames{n};
% Name the new statmap pattern and thresholded mask that will be created
cur_maskname = sprintf('%s_%i',args.new_maskstem,n);
cur_map_patname = sprintf('%s_%i',args.new_map_patname,n);
% if a pattern with the same name already exists, it
% will trigger an error later in init_object, but we
% want to catch it here to save running the entire
% statmap first
if exist_object(subj,'pattern',cur_map_patname)
error('A pattern called %s already exists',cur_map_patname);
end
if ~isempty(args.statmap_arg) && ~isstruct(args.statmap_arg)
warning('Statmap_arg is supposed to be a struct');
end
% Add the current iteration number to the extra_arg, just in case
% it's useful
args.statmap_arg(1).cur_iteration = n;
% Create a handle for the statmap function handle and then run it
% to generate the statmaps
statmap_fh = str2func(args.statmap_funct);
subj = statmap_fh(subj,cur_data_patname,regsname,cur_selname,cur_map_patname,args.statmap_arg);
subj = set_objfield(subj,'pattern',cur_map_patname,'group_name',args.new_map_patname);
if ~isempty(args.thresh)
% Now, create a new thresholded binary mask from the p-values
% statmap pattern returned by the anova
subj = create_thresh_mask(subj,cur_map_patname,cur_maskname,args.thresh);
subj = set_objfield(subj,'mask',cur_maskname,'group_name',args.new_maskstem);
end
end % i nIterations
disp(' ');
disp( sprintf('Pattern statmap group ''%s'' and mask group ''%s'' created by feature_select', ...
args.new_map_patname,args.new_maskstem) );
这是上面函数的一部分，我已将其改编为使用我自己的掩码：
dirIn_Mask = ['/D disk/my_fold/My_generated_Mask'];
data_patin = subj.patterns{1,2}.name;
defaults.new_maskstem = sprintf(‘whole_brain_mask’);
defaults.statmap_arg = struct([]);
args = propval(varargin,defaults);
args.new_maskstem = sprintf( '%s',args.new_maskstem );
for n = 1:1:10
fprintf(' %i', n );
inputFile_mask = ['whole_brain_mask_',num2str(n)];
load(fullfile(dirIn_Mask,inputFile_mask));
cur_maskname = sprintf('%s_%i',args.new_maskstem,n );
subj.masks{1,(runID+1)}.mat = whole_brain_new_1;
subj = set_objfield(subj,'mask',cur_maskname,'group_name',new_maskstem);
end
这是我得到的错误：
Attempt to execute SCRIPT varargin as a function:
C:\D disk\MATLAB\R2014a\toolbox\matlab\lang\varargin.m
Error in MVPA (line 89)
args = propval(varargin,defaults);
谁能帮我解决这个问题？
从您的代码中删除 varargin。 varargin 是一个特殊关键字，仅当您 运行 将代码作为函数时才可用。 varargin 允许将可变数量的输入放入您的函数中。这些输入被散列到一个元胞数组中。
因为这是一个脚本文件，所以那个变量不存在。从您收到的 MATLAB 错误中也可以看出这一点。 FWIW，MATLAB 错误消息非常冗长。很难误解或误解他们想说的话。
因此，这样做：
args = propval(defaults);
如果这不起作用，那么我不确定 propval 是如何接受输入的，但也许可以通过放入一个空元胞数组来愚弄它：
args = propval(cell(1,1), defaults);
因为函数 propval 依赖于可变数量的输入，它应该处理 varargin 不存在的情况，所以我们可以模拟 varargin 没有额外的使用空元胞数组的参数。
您正在使用 varargin 作为 propval 的参数，但 varargin 未在您的代码中的任何位置定义。在这种情况下，varargin 回滚到 Matlab 脚本 varargin。结果，解释器假设你想 运行 它作为一个函数，returns 值到 propval 并且假设脚本不应该 return 值，你得到错误信息。
附带说明：varargin 通常设置为函数的最后一个参数，以允许程序员声明具有可变数量参数的函数。