Breast Cancer - XAI Course

#this method should be added in the class of DiseaseClassifier
def predict_proba(self, x):
    with torch.no_grad():
        x = torch.tensor(x, dtype=torch.float32)
        pred = torch.nn.functional.softmax(self.forward(x).cpu().detach())

        return np.array(pred)
        
        
#LORE 
bbox = sklearn_classifier_wrapper(model)
explainer = LoreTabularExplainer(bbox)

inst_list = X_train[100:101] #list of lists
inst_list[0] #list -- to give it to the explainer

target = "diagnosis"


explainer = LoreTabularExplainer(bbox)
config = {'neigh_type':'rndgen', 'size':1000, 'ocr':0.1, 'ngen':10}
explainer.fit(data, target, config)
exp = explainer.explain(inst_list[0])
print(exp)


exp.plotRules()
exp.plotCounterfactualRules()
#These two lines will give us rules and counterfactual rules.


#LIME
feature_names = data.columns
explainer = lime.lime_tabular.LimeTabularExplainer(X_train, feature_names = feature_names)
exp = explainer.explain_instance(inst_list[0], bbox.predict_proba)
exp.show_in_notebook(show_table=True, show_all=False)