def test_limits():
p = (ggplot(df, aes('x')) +
stat_function(fun=np.cos, size=2,
color='blue', arrow=arrow(ends='first')) +
stat_function(fun=np.cos, xlim=(10, 20), size=2,
color='red', arrow=arrow(ends='last')))
assert p == 'limits'
def test_limits():
p = (ggplot(df, aes('x')) +
stat_function(fun=np.cos, size=2,
color='blue', arrow=arrow(ends='first')) +
stat_function(fun=np.cos, xlim=(10, 20), size=2,
color='red', arrow=arrow(ends='last')))
assert p == 'limits'
def test_exceptions():
# no x limits
with pytest.raises(PlotnineError):
p = ggplot(df)
print(p + stat_function(fun=np.sin))
# fun not callable
with pytest.raises(PlotnineError):
p = ggplot(df, aes('x'))
print(p + stat_function(fun=1))
def test_exceptions():
# no x limits
with pytest.raises(PlotnineError):
p = ggplot(df)
print(p + stat_function(fun=np.sin))
# fun not callable
with pytest.raises(PlotnineError):
p = ggplot(df, aes('x'))
print(p + stat_function(fun=1))
def test_args():
def fun(x, f=lambda x: x, mul=1, add=0):
return f(x)*mul + add
# no args, single arg, tuple of args, dict of args
p = (ggplot(df, aes('x')) +
stat_function(fun=fun, size=2, color='blue') +
stat_function(fun=fun, size=2, color='red', args=np.cos) +
stat_function(fun=fun, size=2, color='green',
args=(np.cos, 2, 1)) +
stat_function(fun=fun, size=2, color='purple',
args=dict(f=np.cos, mul=3, add=2)))
assert p == 'args'
def test_args():
def fun(x, f=lambda x: x, mul=1, add=0):
return f(x)*mul + add
# no args, single arg, tuple of args, dict of args
p = (ggplot(df, aes('x')) +
stat_function(fun=fun, size=2, color='blue') +
stat_function(fun=fun, size=2, color='red', args=np.cos) +
stat_function(fun=fun, size=2, color='green',
args=(np.cos, 2, 1)) +
stat_function(fun=fun, size=2, color='purple',
args=dict(f=np.cos, mul=3, add=2)))
assert p == 'args'
def fit_curve(self):
df = load_protobowl()
# convert prompt to false
df.result = df.result.apply(lambda x: x is True)
xy = list(zip(df.relative_position.tolist(), df.result.tolist()))
xy = sorted(xy, key=lambda x: x[0])
ratios = dict()
cnt = 0
for x, y in xy:
x = int(x * 1000)
ratios[x] = cnt
cnt += y
ratios = sorted(ratios.items(), key=lambda x: x[0])
ratios = [(x / 1000, y) for x, y in ratios]
ttl_correct = df.result.tolist().count(True)
ttl_correct = len(xy)
curve = [(x, 1 - y / ttl_correct) for x, y in ratios]
X, y = list(map(list, zip(*curve)))
X = np.asarray(X)
y = np.asarray(y)
degree = 3
polynomial_features = PolynomialFeatures(degree=degree,
include_bias=False)
linear_regression = LinearRegression()
pipeline = Pipeline([
("polynomial_features", polynomial_features),
("linear_regression", linear_regression),
])
pipeline.fit(X[:, np.newaxis], y)
print(pipeline.steps[1][1].coef_)
def get_weight(x):
return pipeline.predict(np.asarray([[x]]))[0]
ddf = pd.DataFrame({"x": X, "y": y})
p0 = (ggplot(ddf, aes(x="x", y="y")) +
geom_point(size=0.3, color="blue", alpha=0.5, shape="+") +
stat_function(fun=get_weight, color="red", size=2, alpha=0.5) +
labs(x="Position", y="Weight"))
p0.save("output/reporting/curve_score.pdf")
p0.draw()
return pipeline
def fit_curve(self):
df, questions = load_protobowl()
# convert prompt to false
df.result = df.result.apply(lambda x: x is True)
xy = list(zip(df.relative_position.tolist(), df.result.tolist()))
xy = sorted(xy, key=lambda x: x[0])
ratios = dict()
cnt = 0
for x, y in xy:
x = int(x*1000)
ratios[x] = cnt
cnt += y
ratios = sorted(ratios.items(), key=lambda x: x[0])
ratios = [(x / 1000, y) for x, y in ratios]
ttl_correct = df.result.tolist().count(True)
ttl_correct = len(xy)
curve = [(x, 1 - y / ttl_correct) for x, y in ratios]
X, y = list(map(list, zip(*curve)))
X = np.asarray(X)
y = np.asarray(y)
degree = 3
polynomial_features = PolynomialFeatures(degree=degree, include_bias=False)
linear_regression = LinearRegression()
pipeline = Pipeline([("polynomial_features", polynomial_features),
("linear_regression", linear_regression)])
pipeline.fit(X[:, np.newaxis], y)
print(pipeline.steps[1][1].coef_)
def get_weight(x):
return pipeline.predict(np.asarray([[x]]))[0]
ddf = pd.DataFrame({'x': X, 'y': y})
p0 = ggplot(ddf, aes(x='x', y='y')) \
+ geom_point(size=0.3, color='blue', alpha=0.5, shape='+') \
+ stat_function(fun=get_weight, color='red', size=2, alpha=0.5) \
+ labs(x='Position', y='Weight')
p0.save('output/reporting/curve_score.pdf')
p0.draw()
return pipeline