perClass Matlab Toolbox Interface
Table of contents
- 1. Examples and help
- 2. sddata class: Handling data sets
- 2.1. Data subsets
- 2.2. Interactive data visualization
- 2.3. Handling data set labels
- 2.4. Handling data set feature labels
- 2.5. Handling data set properties
- 2.6. Operations on data sets
- 3. sdlab class: Handling labels, decisions and indexed properties
- 4. sdlist class: List of categories
- 5. Feature extraction
- 6. sdppl class: Pipelines and classifiers
- 6.1. Constructing and using pipelines
- 6.2. Dimensionality reduction and data representation
- 6.3. Classifiers (models)
- 6.4. High-level classifier tools
- 7. sdroc class: ROC analysis and evaluation
- 7.1. Performing ROC analysis; accessing operating points and performance estimates
- 7.2. Selecting a specific operating point
- 7.3. Evaluation and performance testing
- 8. sdalg class: Custom algorithms
- 9. sddb class: SQLite database object
- 10. sdsql class: SQLite database query
1. Examples and help ↩
sddemo
- Demo examplessdfeedback
- Send feedback or bug reports to PR Sys Designsdversion
- Display version and license information
2. sddata class: Handling data sets ↩
sddata
- Data set objectsdimport
- Import sddata object from text filesdexport
- Export sddata object into text filegetdata
- Get data matrixsetdata
- Set data matrix (useful for creating a data set with newly computed features preserving all other meta-data)double
- Convert to doublegetname
- Get data set name stringsetname
- Set data set name stringsize
- Get data set size (number of samples, features and classes)sdnominal
- View and handle nominal features
2.1. Data subsets ↩
subset
- Subset of samples given label/property valuesfind
- Get indices of samples by label/property valuesrandsubset
- Random subset sampling classes or any property
2.2. Interactive data visualization ↩
sdscatter
- Interactive scatter plot; hand-painting class labels; filtering samplessdimage
- Image plot with hand-painting of image labels, cropping, connected componentssdfeatplot
- Interactive plot of per-feature class distributions
2.3. Handling data set labels ↩
getlab
- Get current labelssetlab
- Set a current labels by name or find out what is current label setgetsizes
- Get vector of class sizes for current labelsisclass
- Test if class or classes are presentsdrelab
- Rename classes, define meta-classes, extract information from labels with regular expressions
2.4. Handling data set feature labels ↩
getfeatlab
- Get current feature labelssetfeatlab
- Set a property as feature labels or find what property is used as labels
2.5. Handling data set properties ↩
getprop
- Get propertysetprop
- Set (create) propertyrmprop
- Remove property from data setgetproplist
- Get list of properties available in data setisprop
- Test if property is present
2.6. Operations on data sets ↩
abs
- Absolute valuecov
- Compute covariance matrix per classmean
- Compute class mean vectorslog
- Natural logarithm of data matrixlog2
- Base 2 logarithm of data matrixlog10
- Base 10 logarithm of data matrix
3. sdlab class: Handling labels, decisions and indexed properties ↩
sdlab
- Create sdlab objectlength
- Get number of entriesgetlist
orlab.list
- Get the label list object describing categoriesgetnames
- Get string name for each label entrygetindices
- Get index to label list for each label entrygetsizes
orlab.sizes
- Get number of entries per category (i.e. samples per class)lab.fractions
- Get the relative fraction for each categorysdrelab
- Rename classes or define meta-classes
4. sdlist class: List of categories ↩
sdlist
- Create list objectgetnames
- Get category nameslength
- Get number of categoriesind2name
- Convert category index to namename2ind
- Convert name to category indexisname
- Test if a category is present
5. Feature extraction ↩
sdprep
- Preprocessing and spectral indicessdextract
- Extract features- from local image regions
- from each object defined by object labels
- from bands in spectral data
- trasform color space
sdsegment
- Define connected components in an image
6. sdppl class: Pipelines and classifiers ↩
6.1. Constructing and using pipelines ↩
getlab
orp.lab
- Get pipeline labels (feature labels of the output data set)getlist
orp.list
- Get pipeline decision listp.inlab
- Get pipeline input labels (names of features it expects on its input)getoutput
- Return pipeline output type (decision, similarity, distance, ...)sdexe
- Execute pipeline on data using perClass Runtime (in Matlab)sdexport
- Export pipeline for execution out of Matlab using perClass Runtime
6.2. Dimensionality reduction and data representation ↩
sdmissing
- Missing data handling and imputationsdpca
- Principal Component Analysis (PCA)sdlda
- Linear Discriminant Analysis (Fisher projection)sdprox
- Construction of proxmity representation (distances or similarities to prototypes)sdscale
- Data scalingsdnorm
- Normalization of soft-outputssdfeatsel
- Feature selectionsdexpand
- Polynomial feature space expansion
6.3. Classifiers (models) ↩
sdbox
- Bounding box classifiersdnmean
- Nearest mean classifiersdlinear
- Linear discriminant assuming normal densitiessdquadratic
- Quadratic discriminant assuming normal densitiessdgauss
- Gaussian modelsdmixture
- Gaussian mixture automatically estimated number of componentssdfisher
- Fisher linear discriminant (LDA + sdlinear)sdknn
- k-th nearest neighbor classifiersdkmeans
- k-means classifiersdkcentres
- k-centres classifiersdlms
- Least-mean squares classifiersdlogistic
- Logistic regression classifiersdlut
- Lookup table classifiersdparzen
- Parzen classifiersdrandforest
- Random forest classifiersdneural
- Feed-forward and RBF neural networksdnbayes
- Naive Bayes classifiersdtree
- Decision tree classifiersdsvc
- Support vector classifier
6.4. High-level classifier tools ↩
sddetect
- Detector for any model based on ROC analysis or one-class approachsdcascade
- Classifier cascade or hierarchy of classifierssdcombine
- Combine multiple classifiers (soft or crisp combining)sdreject
- Adding reject option to a trained pipelinesdstackgen
- Stacked generalization for building trained combiners (produces unbiased model output on the training set by cross-validation)sdrelab
- Arbitrarily rename decisions of a classifier (also joining decisions)
7. sdroc class: ROC analysis and evaluation ↩
7.1. Performing ROC analysis; accessing operating points and performance estimates ↩
sdroc
- Performing ROC analysis on classifier outputsgetdata
- Get data (performance estimates) stored in the sdroc objectsddrawroc
- Interactive ROC plot
7.2. Selecting a specific operating point ↩
getcurop
- Get index of the current operating pointsetcurop
- Set the current operating point by index or performancesubset
- Select a subset of operating pointsconstrain
- Select a subset of operating point by applying performance constraintssddecide
- Add a default operating point to a trained classifier model
7.3. Evaluation and performance testing ↩
sdtest
- Estimating error/performance of a classifiersdconfmat
- Estimate confusion matrices from decisions or classifer outputsdconfmatind
- Get indices of samples falling into a specific confusion matrix cellsdloss
- Compute loss based on confusion matricessdcrossval
- Cross-validation over samples or a user-defined property
8. sdalg class: Custom algorithms ↩
sdalg
- Construct algorithm objectistrained
- Test if the algorithm is trainedsda_pca_clf
- Algorithm example: PCA dim.reduction followed by a classifiersda_pca_clf_roc
- Algorithm example: PCA dim.reduction + classifier + set operating point by ROC analysis
9. sddb class: SQLite database object ↩
sddb
- Data base object (open data base)tables
- Return tables in a data basefields
- List table fieldsinsert
- Insert data into table
10. sdsql class: SQLite database query ↩
sdsql
- Data base query objectclose
- Close the querydouble
- Convert query results into double matrixcell
- Convert query results into cell array