Configurations - RodentDataAnalytics/mwm-ml-gen GitHub Wiki

1. Options > Figure Configurations

2. Options > Tags Configurations

3. Advanced > Method Configurations

4. Advanced > Classification Configurations

Customise figures properties and exporting.

• FontName: List of available fonts.

• FonSize: List of available text sizes.

• LineWidth: List of available line widths.

• Export: Specifies the export format and quality of the generated figures. MATLAB FIG-file(.fig) can only be opened and processed by MATLAB.

• Default: Sets the default options: Arial font with size 12 and line width of 0.5. Export format is set to JPEG image (.jpg) and quality is set to low.

Customise labels properties and classes of behaviour.

Table columns:

• Abbreviation: the abbreviation of each strategy.

• Name: the name of each strategy.

• ID: the ID of each strategy.

• Color & Linestyle: these options are not currently in use.

• Note: specifies for each strategy if it will be used for the full Trajectories (trajectories) or the Segments (segment) or both (public).

Buttons:

• Edit: Edit the selected strategy. Strategy IDs cannot be changed.

• Default: Sets the default options.

Since the classification has been performed on overlapping segments of the animals' swimming paths we need to map them back to the whole trajectories and computed the evolution of the strategies. This is done by a smoothing function which depends on the arena dimensions and is defined as follows:

where T_i is the i_th interval, d_i,j is the distance from the centre of the j_th segment S_j overlapping the i_th interval to the centre of the i_th, c_k is the k_th segment class and w_k is a class weight normalised so that Σw_k=1. The sum is to be taken over the segments intersecting with the interval T_i, belong to class c_k (unclassified segments are excluded) and fulfil the requirement e^fraction >= 0.14, where for σ equal to R the threshold of 0.14 equals to 2xR. Finally, the class weight w_k was defined as w_k = 1/Pc_k, where Pc_k is the percentage of segments belonging to class k.

This smoothing function can be deactivated and in that case the segments are considered stand alone data; the strategy analysis conclusions should be equivalent with and without the smoothing function, but the transitions will be overestimated and significant difference may be lost on that occasion. In addition the smoothing function can be edited by specifying a different σ and/or path interval (in cm); in case the xR is selected then the cm are tuned into times the arena radius.

Changing the default options for the Strategies apply only to the strategies results while changing the default options for the Transitions apply also to the strategy probabilities results. In addition any folder holding the results generated without the smoothing function will contain the note '_nosmooth'.

Changes the default classification and cross validation options.

Generates and test classifiers: Configures the 10 fold cross validation procedure by specifying what will be the initial and the final number of clusters K that will be tested. If start K equals to 0 then by default the initial K equals the number of labels that have been used increased by 2. The process can also stop if a number of clusters generates error more or equal than a certain threshold. If this threshold is set to 0 then this rule is ignored.

Classifier's Pool: Configures the default classification process by selecting which classifiers should be kept inside the pool based on their cross validation error. The second rule of how many classifiers should be generated inside the pool in order to hold or drop the classification has currently no effect. The reason for this is because this information can be obtained from the confidence intervals of the classifiers which should be clearly above 50% (refer to the Results section Strategies).

Default: Sets the default options.