minor fix (#1086)

* minor fix * minor fix * Process tutorial notebooks --------- Co-authored-by: GitHub Action <[email protected]>
NeuromatchAcademy · Oct 25, 2023 · 061f96b · 061f96b
1 parent 3d638d0
commit 061f96b
Show file tree

Hide file tree

Showing 6 changed files with 34 additions and 34 deletions.
diff --git a/tutorials/W1D3_GeneralizedLinearModels/W1D3_Tutorial2.ipynb b/tutorials/W1D3_GeneralizedLinearModels/W1D3_Tutorial2.ipynb
@@ -205,7 +205,7 @@
     "  best_C = C_values[np.argmax(accuracies)]\n",
     "  ax.set(\n",
     "      xticks=C_values,\n",
-    "      xlabel=\"$C$\",\n",
+    "      xlabel=\"C\",\n",
     "      ylabel=\"Cross-validated accuracy\",\n",
     "      title=f\"Best C: {best_C:1g} ({np.max(accuracies):.2%})\",\n",
     "  )\n",
@@ -219,7 +219,7 @@
     "  ax.plot(C_values, non_zero_l1, marker=\"o\")\n",
     "  ax.set(\n",
     "    xticks=C_values,\n",
-    "    xlabel=\"$C$\",\n",
+    "    xlabel=\"C\",\n",
     "    ylabel=\"Number of non-zero coefficients\",\n",
     "  )\n",
     "  ax.axhline(n_voxels, color=\".1\", linestyle=\":\")\n",
@@ -640,7 +640,7 @@
     "\n",
     "*Estimated timing to here from start of tutorial: 30 min*\n",
     "\n",
-    "Now we need to evaluate the model's predictions. We'll do that with an *accuracy* score. The accuracy of the classifier is the proportion of trials where the predicted label matches the true label.\n"
+    "Now, we need to evaluate the predictions of the model. We will use an *accuracy* score for this purpose. The accuracy of a classifier is determined by the proportion of correct trials, where the predicted label matches the true label, out of the total number of trials."
    ]
   },
   {
@@ -726,7 +726,7 @@
     "\n",
     "  y_pred = model.predict(X)\n",
     "\n",
-    "  accuracy = (y == y_pred).mean()\n",
+    "  accuracy = (y == y_pred).sum() / len(y)\n",
     "\n",
     "  return accuracy\n",
     "\n",
@@ -841,7 +841,7 @@
    },
    "outputs": [],
    "source": [
-    "X.shape"
+    "print(X.shape)"
    ]
   },
   {
@@ -1089,10 +1089,10 @@
     "  penalized_models[log_C] = m.fit(X, y)\n",
     "\n",
     "@widgets.interact\n",
-    "def plot_observed(log_C = widgets.FloatSlider(value=1, min=1, max=10, step=1)):\n",
+    "def plot_observed(log_C = widgets.IntSlider(value=1, min=1, max=10, step=1)):\n",
     "  models = {\n",
     "    \"No regularization\": log_reg,\n",
-    "    f\"$L_2$ (C = $10^{log_C}$)\": penalized_models[log_C]\n",
+    "    f\"$L_2$ (C = $10^{{{log_C}}}$)\": penalized_models[log_C]\n",
     "  }\n",
     "  plot_weights(models)"
    ]
@@ -1103,7 +1103,7 @@
     "execution": {}
    },
    "source": [
-    "Recall from above that $C=\\frac1\\beta$ so larger `C` is less regularization. The top panel corresponds to $C=\\infty$."
+    "Recall from above that $C=\\frac1\\beta$ so larger $C$ is less regularization. The top panel corresponds to $C \\rightarrow \\infty$."
    ]
   },
   {
@@ -1347,7 +1347,7 @@
     "execution": {}
    },
    "source": [
-    "Smaller `C` (bigger $\\beta$) leads to sparser solutions.\n",
+    "Smaller $C$ (bigger $\\beta$) leads to sparser solutions.\n",
     "\n",
     "**Link to neuroscience**: When is it OK to assume that the parameter vector is sparse? Whenever it is true that most features don't affect the outcome. One use-case might be decoding low-level visual features from whole-brain fMRI: we may expect only voxels in V1 and thalamus should be used in the prediction.\n",
     "\n",
@@ -1565,7 +1565,7 @@
     "\n",
     "**The logistic link function**\n",
     "\n",
-    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^T x_i$ to keep it between $0$ and $1$:\n",
+    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^\\top x_i$ to keep it between $0$ and $1$:\n",
     "\n",
     "\\begin{equation}\n",
     "\\sigma(z) = \\frac{1}{1 + \\textrm{exp}(-z)}\n",
@@ -1661,7 +1661,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.17"
+   "version": "3.9.18"
   }
  },
  "nbformat": 4,

diff --git a/tutorials/W1D3_GeneralizedLinearModels/instructor/W1D3_Tutorial2.ipynb b/tutorials/W1D3_GeneralizedLinearModels/instructor/W1D3_Tutorial2.ipynb
@@ -205,7 +205,7 @@
     "  best_C = C_values[np.argmax(accuracies)]\n",
     "  ax.set(\n",
     "      xticks=C_values,\n",
-    "      xlabel=\"$C$\",\n",
+    "      xlabel=\"C\",\n",
     "      ylabel=\"Cross-validated accuracy\",\n",
     "      title=f\"Best C: {best_C:1g} ({np.max(accuracies):.2%})\",\n",
     "  )\n",
@@ -219,7 +219,7 @@
     "  ax.plot(C_values, non_zero_l1, marker=\"o\")\n",
     "  ax.set(\n",
     "    xticks=C_values,\n",
-    "    xlabel=\"$C$\",\n",
+    "    xlabel=\"C\",\n",
     "    ylabel=\"Number of non-zero coefficients\",\n",
     "  )\n",
     "  ax.axhline(n_voxels, color=\".1\", linestyle=\":\")\n",
@@ -642,7 +642,7 @@
     "\n",
     "*Estimated timing to here from start of tutorial: 30 min*\n",
     "\n",
-    "Now we need to evaluate the model's predictions. We'll do that with an *accuracy* score. The accuracy of the classifier is the proportion of trials where the predicted label matches the true label.\n"
+    "Now, we need to evaluate the predictions of the model. We will use an *accuracy* score for this purpose. The accuracy of a classifier is determined by the proportion of correct trials, where the predicted label matches the true label, out of the total number of trials."
    ]
   },
   {
@@ -730,7 +730,7 @@
     "\n",
     "  y_pred = model.predict(X)\n",
     "\n",
-    "  accuracy = (y == y_pred).mean()\n",
+    "  accuracy = (y == y_pred).sum() / len(y)\n",
     "\n",
     "  return accuracy\n",
     "\n",
@@ -845,7 +845,7 @@
    },
    "outputs": [],
    "source": [
-    "X.shape"
+    "print(X.shape)"
    ]
   },
   {
@@ -1093,10 +1093,10 @@
     "  penalized_models[log_C] = m.fit(X, y)\n",
     "\n",
     "@widgets.interact\n",
-    "def plot_observed(log_C = widgets.FloatSlider(value=1, min=1, max=10, step=1)):\n",
+    "def plot_observed(log_C = widgets.IntSlider(value=1, min=1, max=10, step=1)):\n",
     "  models = {\n",
     "    \"No regularization\": log_reg,\n",
-    "    f\"$L_2$ (C = $10^{log_C}$)\": penalized_models[log_C]\n",
+    "    f\"$L_2$ (C = $10^{{{log_C}}}$)\": penalized_models[log_C]\n",
     "  }\n",
     "  plot_weights(models)"
    ]
@@ -1107,7 +1107,7 @@
     "execution": {}
    },
    "source": [
-    "Recall from above that $C=\\frac1\\beta$ so larger `C` is less regularization. The top panel corresponds to $C=\\infty$."
+    "Recall from above that $C=\\frac1\\beta$ so larger $C$ is less regularization. The top panel corresponds to $C \\rightarrow \\infty$."
    ]
   },
   {
@@ -1353,7 +1353,7 @@
     "execution": {}
    },
    "source": [
-    "Smaller `C` (bigger $\\beta$) leads to sparser solutions.\n",
+    "Smaller $C$ (bigger $\\beta$) leads to sparser solutions.\n",
     "\n",
     "**Link to neuroscience**: When is it OK to assume that the parameter vector is sparse? Whenever it is true that most features don't affect the outcome. One use-case might be decoding low-level visual features from whole-brain fMRI: we may expect only voxels in V1 and thalamus should be used in the prediction.\n",
     "\n",
@@ -1573,7 +1573,7 @@
     "\n",
     "**The logistic link function**\n",
     "\n",
-    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^T x_i$ to keep it between $0$ and $1$:\n",
+    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^\\top x_i$ to keep it between $0$ and $1$:\n",
     "\n",
     "\\begin{equation}\n",
     "\\sigma(z) = \\frac{1}{1 + \\textrm{exp}(-z)}\n",
@@ -1669,7 +1669,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.17"
+   "version": "3.9.18"
   }
  },
  "nbformat": 4,

diff --git a/...tions/W1D3_Tutorial2_Solution_bfe654b0.py → ...tions/W1D3_Tutorial2_Solution_1485808c.py b/...tions/W1D3_Tutorial2_Solution_bfe654b0.py → ...tions/W1D3_Tutorial2_Solution_1485808c.py
@@ -13,7 +13,7 @@ def compute_accuracy(X, y, model):
 
   y_pred = model.predict(X)
 
-  accuracy = (y == y_pred).mean()
+  accuracy = (y == y_pred).sum() / len(y)
 
   return accuracy
 

diff --git a/...als/W1D3_GeneralizedLinearModels/static/W1D3_Tutorial2_Solution_19d58990_16.png b/...als/W1D3_GeneralizedLinearModels/static/W1D3_Tutorial2_Solution_19d58990_16.png
diff --git a/...als/W1D3_GeneralizedLinearModels/static/W1D3_Tutorial2_Solution_6bf38e57_32.png b/...als/W1D3_GeneralizedLinearModels/static/W1D3_Tutorial2_Solution_6bf38e57_32.png
diff --git a/tutorials/W1D3_GeneralizedLinearModels/student/W1D3_Tutorial2.ipynb b/tutorials/W1D3_GeneralizedLinearModels/student/W1D3_Tutorial2.ipynb
@@ -205,7 +205,7 @@
     "  best_C = C_values[np.argmax(accuracies)]\n",
     "  ax.set(\n",
     "      xticks=C_values,\n",
-    "      xlabel=\"$C$\",\n",
+    "      xlabel=\"C\",\n",
     "      ylabel=\"Cross-validated accuracy\",\n",
     "      title=f\"Best C: {best_C:1g} ({np.max(accuracies):.2%})\",\n",
     "  )\n",
@@ -219,7 +219,7 @@
     "  ax.plot(C_values, non_zero_l1, marker=\"o\")\n",
     "  ax.set(\n",
     "    xticks=C_values,\n",
-    "    xlabel=\"$C$\",\n",
+    "    xlabel=\"C\",\n",
     "    ylabel=\"Number of non-zero coefficients\",\n",
     "  )\n",
     "  ax.axhline(n_voxels, color=\".1\", linestyle=\":\")\n",
@@ -633,7 +633,7 @@
     "\n",
     "*Estimated timing to here from start of tutorial: 30 min*\n",
     "\n",
-    "Now we need to evaluate the model's predictions. We'll do that with an *accuracy* score. The accuracy of the classifier is the proportion of trials where the predicted label matches the true label.\n"
+    "Now, we need to evaluate the predictions of the model. We will use an *accuracy* score for this purpose. The accuracy of a classifier is determined by the proportion of correct trials, where the predicted label matches the true label, out of the total number of trials."
    ]
   },
   {
@@ -702,7 +702,7 @@
     "execution": {}
    },
    "source": [
-    "[*Click for solution*](https://github.com/NeuromatchAcademy/course-content/tree/main/tutorials/W1D3_GeneralizedLinearModels/solutions/W1D3_Tutorial2_Solution_bfe654b0.py)\n",
+    "[*Click for solution*](https://github.com/NeuromatchAcademy/course-content/tree/main/tutorials/W1D3_GeneralizedLinearModels/solutions/W1D3_Tutorial2_Solution_1485808c.py)\n",
     "\n"
    ]
   },
@@ -811,7 +811,7 @@
    },
    "outputs": [],
    "source": [
-    "X.shape"
+    "print(X.shape)"
    ]
   },
   {
@@ -1059,10 +1059,10 @@
     "  penalized_models[log_C] = m.fit(X, y)\n",
     "\n",
     "@widgets.interact\n",
-    "def plot_observed(log_C = widgets.FloatSlider(value=1, min=1, max=10, step=1)):\n",
+    "def plot_observed(log_C = widgets.IntSlider(value=1, min=1, max=10, step=1)):\n",
     "  models = {\n",
     "    \"No regularization\": log_reg,\n",
-    "    f\"$L_2$ (C = $10^{log_C}$)\": penalized_models[log_C]\n",
+    "    f\"$L_2$ (C = $10^{{{log_C}}}$)\": penalized_models[log_C]\n",
     "  }\n",
     "  plot_weights(models)"
    ]
@@ -1073,7 +1073,7 @@
     "execution": {}
    },
    "source": [
-    "Recall from above that $C=\\frac1\\beta$ so larger `C` is less regularization. The top panel corresponds to $C=\\infty$."
+    "Recall from above that $C=\\frac1\\beta$ so larger $C$ is less regularization. The top panel corresponds to $C \\rightarrow \\infty$."
    ]
   },
   {
@@ -1281,7 +1281,7 @@
     "execution": {}
    },
    "source": [
-    "Smaller `C` (bigger $\\beta$) leads to sparser solutions.\n",
+    "Smaller $C$ (bigger $\\beta$) leads to sparser solutions.\n",
     "\n",
     "**Link to neuroscience**: When is it OK to assume that the parameter vector is sparse? Whenever it is true that most features don't affect the outcome. One use-case might be decoding low-level visual features from whole-brain fMRI: we may expect only voxels in V1 and thalamus should be used in the prediction.\n",
     "\n",
@@ -1465,7 +1465,7 @@
     "\n",
     "**The logistic link function**\n",
     "\n",
-    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^T x_i$ to keep it between $0$ and $1$:\n",
+    "You've seen $\\theta^T x_i$ before, but the $\\sigma$ is new. It's the *sigmoidal* or *logistic* link function that \"squashes\" $\\theta^\\top x_i$ to keep it between $0$ and $1$:\n",
     "\n",
     "\\begin{equation}\n",
     "\\sigma(z) = \\frac{1}{1 + \\textrm{exp}(-z)}\n",
@@ -1561,7 +1561,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.17"
+   "version": "3.9.18"
   }
  },
  "nbformat": 4,