# Do not use packages that are not in standard distribution of python
import numpy as np
from ._base_network import _baseNetwork
class SoftmaxRegression(_baseNetwork):
def __init__(self, input_size=28*28, num_classe
...
# Do not use packages that are not in standard distribution of python
import numpy as np
from ._base_network import _baseNetwork
class SoftmaxRegression(_baseNetwork):
def __init__(self, input_size=28*28, num_classes=10):
'''
A single layer softmax regression. The network is composed by:
a linear layer without bias => (optional ReLU activation) => Softmax
:param input_size: the input dimension
:param num_classes: the number of classes in total
'''
super().__init__(input_size, num_classes)
self._weight_init()
def _weight_init(self):
'''
initialize weights of the single layer regression network. No bias term
included.
:return: None; self.weights is filled based on method
- W1: The weight matrix of the linear layer of shape (num_features,
hidden_size)
'''
np.random.seed(1024)
self.weights['W1'] = 0.001 * np.random.randn(self.input_size,
self.num_classes)
self.gradients['W1'] = np.zeros((self.input_size, self.num_classes))
# self.y_out = np.zeros((self.input_size, self.num_classes))
self.count = 0
def forward(self, X, y, mode='train'):
'''
Compute loss and gradients using softmax with vectorization.
:param X: a batch of image (N, 28x28)
:param y: labels of images in the batch (N,)
:return:
loss: the loss associated with the batch
accuracy: the accuracy of the batch
'''
loss = None
gradient = None
accuracy = None
#############################################################################
# TODO:
#
# 1) Implement the forward process and compute the Cross-Entropy loss
#
# 2) Compute the gradient with respect to the loss
#
# Hint:
#
# Store your intermediate outputs before ReLU for backwards
#
#############################################################################
# np.matmul = Matrix product of two arrays.
model = _baseNetwork()
y_score = np.dot(X, self.weights['W1'])
y_Relu = model.ReLU(y_score)
y_out = model.softmax(y_Relu)
This study source was downloaded by 100000830604181 from CourseHero.com on 09-19-2021 07:08:10 GMT -05:00
https://www.coursehero.com/file/83114285/softmax-regressionpy/
This study resou
[Show More]
