def initialize_svm(): fname = 'source/config.yaml' with open(fname, 'r') as stream: yaml_file = yaml.load(stream) #feature_selection = yaml_file['feature_selection'] grid_search = yaml_file['grid_search'] #svm_cost = yaml_file['svm_cost'] #svm_gamma = yaml_file['svm_gamma'] csv_file = yaml_file['csv_file'] data_frame = pd.read_csv(csv_file) if grid_search.lower() == "false": grid_search = False else: grid_search = True SVM = SVM_Model(data_frame, grid_search) return SVM
lda_m = LDA_Model(CLASS_LABELS) models['lda'] = Model(lda_m) ridge_m = Ridge_Model(CLASS_LABELS) models['ridge'] = Model(ridge_m) ridge_m_10 = Ridge_Model(CLASS_LABELS) ridge_m.lmbda = 10.0 models['ridge_lmda_10'] = Model(ridge_m_10) ridge_m_01 = Ridge_Model(CLASS_LABELS) ridge_m.lmbda = 0.1 models['ridge_lmda_01'] = Model(ridge_m_01) svm_m = SVM_Model(CLASS_LABELS) models['svm'] = Model(svm_m) svm_m_10 = SVM_Model(CLASS_LABELS) svm_m.C = 10.0 models['svm_C_10'] = Model(svm_m_10) svm_m_01 = SVM_Model(CLASS_LABELS) svm_m.C = 0.1 models['svm_C_01'] = Model(svm_m_01) #########GRID SEARCH OVER MODELS############ highest_accuracy = 0 # Highest validation accuracy
model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN LDA REGRESSION##### lda_m = LDA_Model(CLASS_LABELS) model = Model(lda_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN QDA REGRESSION##### qda_m = QDA_Model(CLASS_LABELS) model = Model(qda_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN SVM REGRESSION##### svm_m = SVM_Model(CLASS_LABELS) model = Model(svm_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val)
def lclass(): # Load Training Data and Labels X = list(np.load('little_x_train.npy')) Y = list(np.load('little_y_train.npy')) # Load Validation Data and Labels X_val = list(np.load('little_x_val.npy')) Y_val = list(np.load('little_y_val.npy')) CLASS_LABELS = ['apple', 'banana', 'eggplant'] # Project Data to 200 Dimensions using CCA feat_dim = max(X[0].shape) projections = Projections(feat_dim, CLASS_LABELS) cca_proj, white_cov = projections.cca_projection(X, Y, k=2) X = projections.project(cca_proj, white_cov, X) X_val = projections.project(cca_proj, white_cov, X_val) ####RUN RIDGE REGRESSION##### ridge_m = Ridge_Model(CLASS_LABELS) model = Model(ridge_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN LDA REGRESSION##### lda_m = LDA_Model(CLASS_LABELS) model = Model(lda_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN QDA REGRESSION##### qda_m = QDA_Model(CLASS_LABELS) model = Model(qda_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN SVM REGRESSION##### svm_m = SVM_Model(CLASS_LABELS) model = Model(svm_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val) ####RUN Logistic REGRESSION##### lr_m = Logistic_Model(CLASS_LABELS) model = Model(lr_m) model.train_model(X, Y) model.test_model(X, Y) model.test_model(X_val, Y_val)