Spaces:
Sleeping
Sleeping
MilesCranmer
commited on
Commit
•
0bf77e2
1
Parent(s):
4582e28
Autogenerate docstring for PySR Regressor
Browse files- pysr/sklearn.py +36 -2
pysr/sklearn.py
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
from pysr import pysr, best_row
|
2 |
from sklearn.base import BaseEstimator
|
|
|
3 |
|
4 |
|
5 |
class PySRRegressor(BaseEstimator):
|
@@ -42,12 +43,28 @@ class PySRRegressor(BaseEstimator):
|
|
42 |
else:
|
43 |
raise NotImplementedError
|
44 |
|
45 |
-
def fit(self, X, y, weights=None):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
self.equations = pysr(
|
47 |
X=X,
|
48 |
y=y,
|
49 |
weights=weights,
|
50 |
-
|
|
|
51 |
)
|
52 |
return self
|
53 |
|
@@ -56,3 +73,20 @@ class PySRRegressor(BaseEstimator):
|
|
56 |
np_format = equation_row["lambda_format"]
|
57 |
|
58 |
return np_format(X)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from pysr import pysr, best_row
|
2 |
from sklearn.base import BaseEstimator
|
3 |
+
import inspect
|
4 |
|
5 |
|
6 |
class PySRRegressor(BaseEstimator):
|
|
|
43 |
else:
|
44 |
raise NotImplementedError
|
45 |
|
46 |
+
def fit(self, X, y, weights=None, variable_names=None):
|
47 |
+
"""Search for equations to fit the dataset.
|
48 |
+
|
49 |
+
:param X: 2D array. Rows are examples, columns are features. If pandas DataFrame, the columns are used for variable names (so make sure they don't contain spaces).
|
50 |
+
:type X: np.ndarray/pandas.DataFrame
|
51 |
+
:param y: 1D array (rows are examples) or 2D array (rows are examples, columns are outputs). Putting in a 2D array will trigger a search for equations for each feature of y.
|
52 |
+
:type y: np.ndarray
|
53 |
+
:param weights: Optional. Same shape as y. Each element is how to weight the mean-square-error loss for that particular element of y.
|
54 |
+
:type weights: np.ndarray
|
55 |
+
:param variable_names: a list of names for the variables, other than "x0", "x1", etc.
|
56 |
+
:type variable_names: list
|
57 |
+
"""
|
58 |
+
if variable_names is None:
|
59 |
+
if "variable_names" in self.params:
|
60 |
+
variable_names = self.params["variable_names"]
|
61 |
+
|
62 |
self.equations = pysr(
|
63 |
X=X,
|
64 |
y=y,
|
65 |
weights=weights,
|
66 |
+
variable_names=variable_names,
|
67 |
+
**{k: v for k, v in self.params.items() if k != "variable_names"},
|
68 |
)
|
69 |
return self
|
70 |
|
|
|
73 |
np_format = equation_row["lambda_format"]
|
74 |
|
75 |
return np_format(X)
|
76 |
+
|
77 |
+
|
78 |
+
# Add the docs from pysr() to PySRRegressor():
|
79 |
+
|
80 |
+
_pysr_docstring_split = []
|
81 |
+
_start_recording = False
|
82 |
+
for line in inspect.getdoc(pysr).split("\n"):
|
83 |
+
# Skip docs on "X" and "y"
|
84 |
+
if ":param binary_operators:" in line:
|
85 |
+
_start_recording = True
|
86 |
+
if ":returns:" in line:
|
87 |
+
_start_recording = False
|
88 |
+
if _start_recording:
|
89 |
+
_pysr_docstring_split.append(line)
|
90 |
+
_pysr_docstring = "\n\t".join(_pysr_docstring_split)
|
91 |
+
|
92 |
+
PySRRegressor.__init__.__doc__ += _pysr_docstring
|