refactor: code de-dup

layers.py

@@ -0,0 +1,29 @@
+import numpy as np
+import types
+from utils import regularize
+
+
+class NNLayer:
+    def __init__(self,
+                 in_size: int,
+                 out_size: int,
+                 lr: float,
+                 activation_func: types.FunctionType):
+        self.W = np.random.uniform(-1, 1, (in_size, out_size))
+        self.B = np.zeros((out_size))
+        self.lr = lr
+        self.last_input = None
+        self.last_output = None
+        self.activation_func = activation_func
+
+    def forward(self, V: np.ndarray) -> np.ndarray:
+        self.last_input = V
+        res = V @ self.W + self.B
+        self.last_output = regularize(self.activation_func(res))
+        return self.last_output
+
+    def backprop(self, error: np.ndarray):
+        dW = np.outer(self.last_input, error)
+        self.W -= self.lr * dW
+        self.B -= self.lr * error
+        return error @ self.W.T