README.md

ML_Project_2_Team_6

The goal of this project is to do Classification of Patterns with the following algorithms:

1. Relevance Vector Machines (RVM)
2. Support Vector Machines (SVM)
3. Gaussian Process Regression (GPR)

Prerequisites:

1. scipy.io
2. numpy
3. sklearn.model_selection -> KFold
4. sklearn.model_selection -> train_test_split
5. sklearn.gaussian_process -> GaussianProcessClassifier
6. skrvm -> RVC (https://github.com/JamesRitchie/scikit-rvm.git)
7. sklearn.svm -> SVC
8. sklearn.gaussian_process.kernels -> RBF
9. time -> time
10. sklearn.decomposition -> PCA
11. sklearn.metrics -> classification_report
12. sklearn.metrics -> confusion_matrix
13. sklearn.metrics -> accuracy_score
14. pandas -> pd

Following are the methods:

1. def TrainMyClassifier(XEstimate, YEstimate, XValidate, TrainMyClassifierParameters)

Input Parameters:

XEstimate: This is the input data of the training set.

YEstimate: This is the labels of the XEstimate(1-5 in our case)

XValidate: This is the input data of the validation set

TrainMyClassifierParameters: This is an array.

Example:
  
  1. For SVM:
      TrainMyClassifierParameters[0] = {
        'C' : 1e5,
        'gamma' : 0.3
      }
      TrainMyClassifierParameters[1] = 'SVM'
  
  2. For RVM:
    TrainMyClassifierParameters[0] = {
        'alpha' : 1e-04,
        'beta' : 1e-04
    }
      TrainMyClassifierParameters[1] = 'RVM'     
  
  3. For GP:
      TrainMyClassifierParameters[0] = {
        'length_scale' : 10
      }
      TrainMyClassifierParameters[1] = 'GP'      

Return Values:

y_pred: Class labels on the validation set

scores: Mean accuracy on the given test data and labels for this particular test prediction

params: Trained Model for the required algorithm. Contains the estimated parameters and hyperparameters within the model. 

2. def MyCrossValidate(XTrain,YTrain2,Nf,Algorithm)

Input Parameters:

XTrain: Input of the training data

YTrain2: Labels of the training data XTrain

Nf: Number of folds for cross-validation.(Nf=5)

Algorithm: 'SVM'/'RVM'/'GP'

Return Values:

YTrain: The class labels for each validation sample

EstParameters: Array of Estimated Parameters (Trained Model) for each validation sample

EstConfMatrices: Array of Confusion Matrices for each validation sample

ConfMatrix: The overall Confusion Matrix

3. def MyConfusionMatrix(Y,YValidate,ClassNames)

Input Parameters:

Y: Labels predicted by our model

YValidate: Actual labels

ClassNames: Array of names of every class. Example in our case: ['One','Two','Three','Four','Five']

Return Values:

c_m: Classfication prediction in function of groundtruth as a fraction of actual class

df: Same as c_m but formated in report ready form

a_s: Returns accuracy classification score for groundtruth labels vs predicted labels for the overall 

4. def TestMyClassifier(XTest, Parameters, EstParameters)

Input Parameters:

XTest: Test input data

Parameters: Array of estimated parameters from the trained model

EstParameters: Array of estimated hyperparameters from the trained model

Return Values:

Ytest: Array of class labels of test data as predicted by these models.