diff --git a/HW_2_sp23_EE374N-385J_Neural_Eng_EMGanalysis/HW2_EMG_Analysis.ipynb b/HW_2_sp23_EE374N-385J_Neural_Eng_EMGanalysis/HW2_EMG_Analysis.ipynb
index 93a27ce..a42126d 100644
--- a/HW_2_sp23_EE374N-385J_Neural_Eng_EMGanalysis/HW2_EMG_Analysis.ipynb
+++ b/HW_2_sp23_EE374N-385J_Neural_Eng_EMGanalysis/HW2_EMG_Analysis.ipynb
@@ -1730,6 +1730,63 @@
     "print(f\"  Train S2 → Test S1 Run 6  :  {acc_s2_to_s1:.4f}  ({acc_s2_to_s1*100:.1f}%)  [+{(acc_s2_to_s1 - chance)*100:.1f} pp above chance]\")\n",
     "print(f\"  Mean transfer accuracy     :  {(acc_s1_to_s2 + acc_s2_to_s1) / 2:.4f}  ({(acc_s1_to_s2 + acc_s2_to_s1) / 2 * 100:.1f}%)\")"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 54,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "── Cross-Subject CNN Transfer Decoding ─────────────────────────────────\n",
+      "  Window: 250 ms (128 samples)  |  Overlap: 75%\n",
+      "  Chance level : 33.3%\n",
+      "\n",
+      "  Train S1 → Test S2 Run 6  :  0.4396  (44.0%)  [+10.6 pp above chance]\n",
+      "  Train S2 → Test S1 Run 6  :  0.4189  (41.9%)  [+8.6 pp above chance]\n",
+      "  Mean transfer accuracy     :  0.4293  (42.9%)\n"
+     ]
+    }
+   ],
+   "source": [
+    "# ── Cross-subject transfer decoding: CNN trained on one subject, tested on the other ──\n",
+    "\n",
+    "WIN_TEST = 250   # ms  (best window for Subject 1 in Part B CV)\n",
+    "OVL_TEST = 0.75\n",
+    "\n",
+    "s1_runs = filtered_data[0]\n",
+    "s2_runs = filtered_data[1]\n",
+    "\n",
+    "# ── Build raw training/test sets for each subject ──────────────────────────────\n",
+    "run_raw1 = [extract_run_Xy(r, WIN_TEST, OVL_TEST, raw=True) for r in s1_runs[:5]]\n",
+    "X_train1_raw = np.vstack([xy[0] for xy in run_raw1])\n",
+    "y_train1_raw = np.concatenate([xy[1] for xy in run_raw1])\n",
+    "\n",
+    "run_raw2 = [extract_run_Xy(r, WIN_TEST, OVL_TEST, raw=True) for r in s2_runs[:5]]\n",
+    "X_train2_raw = np.vstack([xy[0] for xy in run_raw2])\n",
+    "y_train2_raw = np.concatenate([xy[1] for xy in run_raw2])\n",
+    "\n",
+    "X_test1_raw, y_test1_raw = extract_run_Xy(s1_runs[5], WIN_TEST, OVL_TEST, raw=True)\n",
+    "X_test2_raw, y_test2_raw = extract_run_Xy(s2_runs[5], WIN_TEST, OVL_TEST, raw=True)\n",
+    "\n",
+    "# ── Train S1 → Test S2 ────────────────────────────────────────────────────────\n",
+    "acc_s1_to_s2 = train_cnn_raw(X_train1_raw, y_train1_raw, X_test2_raw, y_test2_raw)\n",
+    "\n",
+    "# ── Train S2 → Test S1 ────────────────────────────────────────────────────────\n",
+    "acc_s2_to_s1 = train_cnn_raw(X_train2_raw, y_train2_raw, X_test1_raw, y_test1_raw)\n",
+    "\n",
+    "chance = 1 / 3\n",
+    "\n",
+    "print(\"── Cross-Subject CNN Transfer Decoding ─────────────────────────────────\")\n",
+    "print(f\"  Window: {WIN_TEST} ms ({ms_to_samples(WIN_TEST)} samples)  |  Overlap: {OVL_TEST:.0%}\")\n",
+    "print(f\"  Chance level : {chance:.1%}\")\n",
+    "print()\n",
+    "print(f\"  Train S1 → Test S2 Run 6  :  {acc_s1_to_s2:.4f}  ({acc_s1_to_s2*100:.1f}%)  [+{(acc_s1_to_s2 - chance)*100:.1f} pp above chance]\")\n",
+    "print(f\"  Train S2 → Test S1 Run 6  :  {acc_s2_to_s1:.4f}  ({acc_s2_to_s1*100:.1f}%)  [+{(acc_s2_to_s1 - chance)*100:.1f} pp above chance]\")\n",
+    "print(f\"  Mean transfer accuracy     :  {(acc_s1_to_s2 + acc_s2_to_s1) / 2:.4f}  ({(acc_s1_to_s2 + acc_s2_to_s1) / 2 * 100:.1f}%)\")"
+   ]
   }
  ],
  "metadata": {