def port_svm(clf, **kwargs):
"""Port a SVC / OneClassSVC classifier"""
assert isinstance(clf.gamma, float), 'You probably didn\'t set an explicit value for gamma: 0.001 is a good default'
support_v = clf.support_vectors_
n_classes = len(clf.n_support_)
return jinja('svm/svm.jinja', {
'kernel': {
'type': clf.kernel,
'gamma': clf.gamma,
'coef0': clf.coef0,
'degree': clf.degree
},
'sizes': {
'features': len(support_v[0]),
'vectors': len(support_v),
'classes': n_classes,
'decisions': n_classes * (n_classes - 1) // 2,
'supports': clf.n_support_
},
'arrays': {
'supports': support_v,
'intercepts': clf.intercept_,
'coefs': clf.dual_coef_
}
}, {
'classname': 'OneClassSVM' if check_type(clf, 'OneClassSVM') else 'SVM'
}, **kwargs)
def port_rvm(clf, classname, **kwargs):
"""Port a RVM classifier"""
assert classname is None or len(classname) > 0, 'Invalid class name'
template_data = {
**kwargs,
'kernel': {
'type': clf.kernel,
'gamma': clf.gamma,
'coef0': clf.coef0,
'degree': clf.degree
},
'sizes': {
'features': len(clf.relevant_vectors_[0]),
},
'arrays': {
'vectors': clf.relevant_vectors_,
'coefs': clf.coef_,
'actives': clf.active_,
'intercepts': clf.intercept_,
'mean': clf._x_mean,
'std': clf._x_std
},
'classname': classname if classname is not None else 'RVM',
}
return jinja('rvm/rvm.jinja', template_data)
def port_svm(clf, classname=None, **kwargs):
"""Port a SVC / LinearSVC classifier"""
assert isinstance(
clf.gamma, float
), 'You probably didn\'t set an explicit value for gamma: 0.001 is a good default'
assert classname is None or len(classname) > 0, 'Invalid class name'
if classname is None:
classname = 'OneClassSVM' if isinstance(clf, OneClassSVM) else 'SVM'
support_v = clf.support_vectors_
n_classes = len(clf.n_support_)
template_data = {
**kwargs, 'kernel': {
'type': clf.kernel,
'gamma': clf.gamma,
'coef0': clf.coef0,
'degree': clf.degree
},
'sizes': {
'features': len(support_v[0]),
'vectors': len(support_v),
'classes': n_classes,
'decisions': n_classes * (n_classes - 1) // 2,
'supports': clf.n_support_
},
'arrays': {
'supports': support_v,
'intercepts': clf.intercept_,
'coefs': clf.dual_coef_
},
'classname': classname
}
return jinja('svm/svm.jinja', template_data)
def port_sefr(clf, classname=None, **kwargs):
kwargs.update({
'weights': clf.weights,
'bias': clf.bias,
'classname': classname or 'SEFR'
})
return jinja('sefr/sefr.jinja', kwargs)
def port_sefr(clf, classname=None, **kwargs):
"""Port SEFR classifier"""
return jinja('sefr/sefr.jinja', {
'weights': clf.weights,
'bias': clf.bias,
'dimension': len(clf.weights),
}, {'classname': 'SEFR'}, **kwargs)
def port_logisticregression(clf, **kwargs):
"""Port sklearn's LogisticRegressionClassifier"""
return jinja(
'logisticregression/logisticregression.jinja', {
'weights': clf.coef_,
'intercept': clf.intercept_,
'classes': clf.classes_,
'n_classes': len(clf.classes_)
}, {'classname': 'LogisticRegression'}, **kwargs)
def port_pca(pca, classname=None, **kwargs):
"""Port a PCA"""
template_data = {
'arrays': {
'components': pca.components_,
'mean': pca.mean_
},
'classname': classname if classname is not None else 'PCA'
}
return jinja('pca/pca.jinja', template_data)
def port_pca(clf, **kwargs):
"""Port a PCA"""
return jinja('pca/pca.jinja', {
'arrays': {
'components': clf.components_,
'mean': clf.mean_
},
}, {
'classname': 'PCA'
}, **kwargs)
def port_linear_regression(clf, classname=None, **kwargs):
"""Port Linear Regression"""
return jinja('linearregression/linearregression.jinja', {
'coefs': clf.coef_,
'intercept': clf.intercept_,
'dimension': len(clf.coef_),
'dtype': 'float'
}, {
'classname': 'LinearRegression'
}, **kwargs)
def port_gaussiannb(clf, **kwargs):
"""Port sklearn's GaussianNB"""
return jinja('gaussiannb/gaussiannb.jinja', {
'sigma': clf.sigma_,
'theta': clf.theta_,
'prior': clf.class_prior_,
'classes': clf.classes_,
'n_classes': len(clf.classes_)
}, {
'classname': 'GaussianNB'
}, **kwargs)
def port_wifi_indoor_positioning(samples):
classmap = get_classmap(samples)
networkmap = get_networkmap(samples)
X = get_x(samples, networkmap)
y = get_y(samples, classmap)
# classmap is flipped wrt the format `port` expects: flip it
classmap = {v: k for k, v in classmap.items()}
return X, y, classmap, jinja('wifiindoorpositioning/wifiindoorpositioning.jinja', {
'X': X,
'networkmap': networkmap
})
def port_decisiontree(clf, **kwargs):
"""Port sklearn's DecisionTreeClassifier"""
return jinja('decisiontree/decisiontree.jinja', {
'left': clf.tree_.children_left,
'right': clf.tree_.children_right,
'features': clf.tree_.feature,
'thresholds': clf.tree_.threshold,
'classes': clf.tree_.value,
'i': 0
}, {
'classname': 'DecisionTree'
}, **kwargs)
def port_randomforest(clf, **kwargs):
"""Port sklearn's RandomForestClassifier"""
return jinja(
'randomforest/randomforest.jinja', {
'n_classes':
clf.n_classes_,
'trees': [{
'left': clf.tree_.children_left,
'right': clf.tree_.children_right,
'features': clf.tree_.feature,
'thresholds': clf.tree_.threshold,
'classes': clf.tree_.value,
} for clf in clf.estimators_]
}, {'classname': 'RandomForest'}, **kwargs)
def port_principalfft(clf,
optimize_sin=False,
lookup_cos=None,
lookup_sin=None,
**kwargs):
"""Port PrincipalFFT classifier"""
return jinja(
'principalfft/principalfft.jinja', {
'fft': clf,
'PI': pi,
'size': len(clf.idx),
'optmize_sin': optimize_sin,
'lookup_cos': lookup_cos,
'lookup_sin': lookup_sin,
}, **kwargs)
def port_xgboost(clf, **kwargs):
"""Port a XGBoost classifier"""
with NamedTemporaryFile('w+', suffix='.json', encoding='utf-8') as tmp:
clf.save_model(tmp.name)
tmp.seek(0)
decoded = json.load(tmp)
trees = [
format_tree(tree) for tree in decoded['learner']
['gradient_booster']['model']['trees']
]
return jinja(
'xgboost/xgboost.jinja', {
'n_classes':
int(decoded['learner']['learner_model_param']['num_class']),
'trees':
trees,
}, {'classname': 'XGBClassifier'}, **kwargs)
def port_rvm(clf, **kwargs):
"""Port a RVM classifier"""
return jinja(
'rvm/rvm.jinja', {
'n_classes': len(clf.intercept_),
'kernel': {
'type': clf.kernel,
'gamma': clf.gamma,
'coef0': clf.coef0,
'degree': clf.degree
},
'sizes': {
'features': clf.relevant_vectors_[0].shape[1],
},
'arrays': {
'vectors': clf.relevant_vectors_,
'coefs': clf.coef_,
'actives': clf.active_,
'intercepts': clf.intercept_,
'mean': clf._x_mean,
'std': clf._x_std
},
}, {'classname': 'RVC'}, **kwargs)