Spaces:
Running
Running
MilesCranmer
commited on
Commit
•
ce5b119
1
Parent(s):
775c667
Add test for feature selection in JAX output
Browse files- test/test_jax.py +19 -0
test/test_jax.py
CHANGED
@@ -5,6 +5,7 @@ import pandas as pd
|
|
5 |
from jax import numpy as jnp
|
6 |
from jax import random
|
7 |
import sympy
|
|
|
8 |
|
9 |
|
10 |
class TestJAX(unittest.TestCase):
|
@@ -79,3 +80,21 @@ class TestJAX(unittest.TestCase):
|
|
79 |
np.square(np.cos(X[:, 1])), # Select feature 1
|
80 |
decimal=4,
|
81 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
from jax import numpy as jnp
|
6 |
from jax import random
|
7 |
import sympy
|
8 |
+
from functools import partial
|
9 |
|
10 |
|
11 |
class TestJAX(unittest.TestCase):
|
|
|
80 |
np.square(np.cos(X[:, 1])), # Select feature 1
|
81 |
decimal=4,
|
82 |
)
|
83 |
+
|
84 |
+
def test_feature_selection(self):
|
85 |
+
X = pd.DataFrame({f"k{i}": np.random.randn(1000) for i in range(10, 21)})
|
86 |
+
y = X["k15"] ** 2 + np.cos(X["k20"])
|
87 |
+
|
88 |
+
model = PySRRegressor(
|
89 |
+
unary_operators=["cos"], select_k_features=3, early_stop_condition=1e-5
|
90 |
+
)
|
91 |
+
model.fit(X.values, y.values)
|
92 |
+
f, parameters = model.jax().values()
|
93 |
+
|
94 |
+
np_prediction = model.predict
|
95 |
+
jax_prediction = partial(f, parameters=parameters)
|
96 |
+
|
97 |
+
np_output = np_prediction(X.values)
|
98 |
+
jax_output = jax_prediction(X.values)
|
99 |
+
|
100 |
+
np.testing.assert_almost_equal(np_output, jax_output, decimal=4)
|