the previous noise function generated a negative when setting the noi…

…se_level to 1. (since this was the flip probability). This implementation generates a fully random pattern on noise_level 1.

Scripts/Cog Neuro Labs/Lab1_2021_DynamicsAndPerception_v1.ipynb

@@ -391,10 +391,22 @@
     "        Xs[i] = X\n",
     "    return Xs\n",
     "\n",
-    "def add_noise(x_, noise_level=.2):\n",
-    "    noise = np.random.choice(\n",
-    "        [1, -1], size=len(x_), p=[1-noise_level, noise_level])\n",
-    "    return x_ * noise\n",
+    "def add_noise(pattern, noise_level=0.2):\n",
+    "    \"\"\"\n",
+    "    Add noise to a `pattern` by replacing each pixel with a random value (-1 or 1)\n",
+    "    with probability `noise_level`\n",
+    "    \"\"\"\n",
+    "    # Generate a random pattern of 1 and -1\n",
+    "    random_pattern = np.random.choice([-1, 1], size=pattern.shape)\n",
+    "\n",
+    "    # Generate a noise mask based on the noise level\n",
+    "    noise_mask = np.random.rand(pattern.shape[0]) < noise_level\n",
+    "\n",
+    "    # Blend the original pattern and the random pattern based on the noise mask\n",
+    "    noisy_pattern = pattern.copy()\n",
+    "    noisy_pattern[noise_mask] = random_pattern[noise_mask]\n",
+    "\n",
+    "    return noisy_pattern\n",
     "\n",
     " \n"
    ],