From 899d0fadd09f67717cfa9200b0c97b53642da129 Mon Sep 17 00:00:00 2001
From: Sean Fuller <54780092+sfuller14@users.noreply.github.com>
Date: Wed, 18 Jan 2023 21:46:26 -0800
Subject: [PATCH] Correct loss calc from SSE to MSE in backprop loop

SSE is used for loss rather than MSE (~1:54:30 in the video lecture)
Using len(ys) for n -- MSE = SSE / n
---
 lectures/micrograd/micrograd_lecture_second_half_roughly.ipynb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lectures/micrograd/micrograd_lecture_second_half_roughly.ipynb b/lectures/micrograd/micrograd_lecture_second_half_roughly.ipynb
index 67a25a68..67a2cb41 100644
--- a/lectures/micrograd/micrograd_lecture_second_half_roughly.ipynb
+++ b/lectures/micrograd/micrograd_lecture_second_half_roughly.ipynb
@@ -1203,7 +1203,7 @@
     "  \n",
     "  # forward pass\n",
     "  ypred = [n(x) for x in xs]\n",
-    "  loss = sum((yout - ygt)**2 for ygt, yout in zip(ys, ypred))\n",
+    "  loss = sum((yout - ygt)**2 for ygt, yout in zip(ys, ypred)) / len(ys) # Video lecture calculates loss as SSE. Corrected to MSE here.\n",
     "  \n",
     "  # backward pass\n",
     "  for p in n.parameters():\n",