first commit

Author: tA-bot-git

Assignment2.ipynb

@@ -0,0 +1,3 @@
+from svm_loss import svm_loss
+from svm_gradient import svm_gradient
+from svm_solver import svm_solver

__init__.py

@@ -0,0 +1,3 @@
+from svm_loss import svm_loss
+from svm_gradient import svm_gradient
+from svm_solver import svm_solver

__pycache__/__init__.cpython-36.pyc

@@ -0,0 +1,36 @@
+import numpy as np
+
+
+def svm_gradient(w, b, x, y, C):
+    """
+    Compute gradient for SVM w.r.t. to the parameters w and b on a mini-batch (x, y)
+
+    Args:
+        w: Parameters of shape [num_features]
+        b: Bias (a scalar)
+        x: A mini-batch of training example [k, num_features]
+        y: Labels corresponding to x of size [k]
+
+    Returns:
+        grad_w: The gradient of the SVM objective w.r.t. w of shape [k, num_features]
+        grad_v: The gradient of the SVM objective w.r.t. b of shape [k, 1]
+
+    """
+
+    grad_w = 0
+    grad_b = 0
+
+
+    #######################################################################
+    # TODO:                                                               #
+    # Compute the gradient for a particular choice of w and b.            #
+    # Compute the partial derivatives and set grad_w and grad_b to the    #
+    # partial derivatives of the cost w.r.t. both parameters              #
+    #                                                                     #
+    #######################################################################
+
+
+    #######################################################################
+    #                         END OF YOUR CODE                            #
+    #######################################################################
+    return grad_w, grad_b