wehub-resource-sync
49 lines
1.1 KiB
Python
Executable File
49 lines
1.1 KiB
Python
Executable File
import numpy as np
|
|
from prml.linear.classifier import Classifier
|
|
from prml.preprocess.label_transformer import LabelTransformer
|
|
|
|
|
|
class LeastSquaresClassifier(Classifier):
|
|
"""
|
|
Least squares classifier model
|
|
|
|
X : (N, D)
|
|
W : (D, K)
|
|
y = argmax_k X @ W
|
|
"""
|
|
|
|
def __init__(self, W:np.ndarray=None):
|
|
self.W = W
|
|
|
|
def fit(self, X:np.ndarray, t:np.ndarray):
|
|
"""
|
|
least squares fitting for classification
|
|
|
|
Parameters
|
|
----------
|
|
X : (N, D) np.ndarray
|
|
training independent variable
|
|
t : (N,) or (N, K) np.ndarray
|
|
training dependent variable
|
|
in class index (N,) or one-of-k coding (N,K)
|
|
"""
|
|
if t.ndim == 1:
|
|
t = LabelTransformer().encode(t)
|
|
self.W = np.linalg.pinv(X) @ t
|
|
|
|
def classify(self, X:np.ndarray):
|
|
"""
|
|
classify input data
|
|
|
|
Parameters
|
|
----------
|
|
X : (N, D) np.ndarray
|
|
independent variable to be classified
|
|
|
|
Returns
|
|
-------
|
|
(N,) np.ndarray
|
|
class index for each input
|
|
"""
|
|
return np.argmax(X @ self.W, axis=-1)
|