Documentation

The name of the class.

The feature vector

Each of the feature is a double value, the interpretation will be based on the distribution type to be used.

just better readability

data set, i.e. the input. each of the feature vector is a point and the classes contain the corresponding class the particular feature belongs to. The size of the points and the classes should be same

Probe is basically a function that gives the probability of a feature having a particular value for a particular className

this is same feature for different input points

this is the parallel data for classes corresponding to the input

Collection of feature vectors

specifies the type of the distribution used to fit the training data for a particular feature Source of Description : MathWork Normal - Gaussian Mvmn - Multivariate multinomial distribution Mn - Multinomial distribution

This defines which prior probablity to use. Empirical - Depends on the training data. Uniform - Equal probablity for all classes Specific - User provided probability

data point, is the combination of point and its class data DataPoint = DataPoint { point::FeatureVector ,className :: ClassName }

alternative to DataSet but in this classes or points are not directly separable

distance measure type DType = String

Resolution method

The number of neighbours the KNN uses

ROMIL

NITESH

General definition of Dataset to unify the interface