remove the emojis

Author: AlexandreAndr

notebooks/Notebook_2/Notebook_2b_Model_Selection.ipynb

@@ -327,8 +327,7 @@
     "instead of trying everything,\n",
     "we **follow the slope of the loss function**.\n",
     "\n",
-    "This leads to…\n",
-    "👉 Gradient Descent\n"
+    "This leads to **Gradient Descent**.\n"
    ]
   },
   {
@@ -4897,7 +4896,7 @@
     "id": "fz4kIYpnASx4"
    },
    "source": [
-    "### ✏️ Additional Questions"
+    "### Additional Questions"
    ]
   },
   {
@@ -4927,16 +4926,16 @@
    "source": [
     "**Challenge 2.**\n",
     "\n",
-    "1️⃣ Why is a model that perfectly fits training data *not always good*?\n",
+    "1. Why is a model that perfectly fits training data *not always good*?\n",
     "\n",
-    "2️⃣ Where do you see “overfitting” in real life?\n",
+    "2. Where do you see “overfitting” in real life?\n",
     "Examples:\n",
     "\n",
     "- memorizing exam answers vs understanding\n",
     "- overtraining in sports\n",
     "- overly tuned rules or policies\n",
     "\n",
-    "3️⃣ Why can’t we judge a model using only training error?\n"
+    "3. Why can’t we judge a model using only training error?\n"
    ]
   },
   {

notebooks/Notebook_2/Notebook_2b_Model_Selection.py

@@ -239,8 +239,7 @@ def plot_grid_search_over_time_gif(
 # instead of trying everything,
 # we **follow the slope of the loss function**.
 #
-# This leads to…
-# 👉 Gradient Descent
+# This leads to **Gradient Descent**.
 #
 
 # %% [markdown] id="vs3Yvhn04l6R"
@@ -1307,7 +1306,7 @@ def validation_curve(max_degree=15):
 #
 
 # %% [markdown] id="fz4kIYpnASx4"
-# ### ✏️ Additional Questions
+# ### Additional Questions
 
 # %% [markdown] id="mPeGKp8BXdm7"
 # **Challenge 1.**
@@ -1323,16 +1322,16 @@ def validation_curve(max_degree=15):
 # %% [markdown] id="79-Rqt9v9wfA"
 # **Challenge 2.**
 #
-# 1️⃣ Why is a model that perfectly fits training data *not always good*?
+# 1. Why is a model that perfectly fits training data *not always good*?
 #
-# 2️⃣ Where do you see “overfitting” in real life?
+# 2. Where do you see “overfitting” in real life?
 # Examples:
 #
 # - memorizing exam answers vs understanding
 # - overtraining in sports
 # - overly tuned rules or policies
 #
-# 3️⃣ Why can’t we judge a model using only training error?
+# 3. Why can’t we judge a model using only training error?
 #
 
 # %% [markdown] id="VP9xE6MYAxeS"

notebooks/Notebook_3/Notebook_3b_Statistics_1.ipynb

@@ -482,7 +482,7 @@
     "\n",
     "**Back to the penguins.** Suppose our data give a 95% confidence interval of $[750\\text{ g},\\ 1050\\text{ g}]$ for the difference in mean body mass between Gentoo and Adelie penguins. This says that the values between 750 g and 1050 g are all plausible estimates of the true difference.\n",
     "\n",
-    "> 💡 **CIs and p-values are two views of the same evidence.** A 95% confidence interval that excludes the null value corresponds to a p-value below $\\alpha = 0.05$ (indeed, $95\\%=0.95=1-0.05=1-\\alpha)$). The confidence interval is often more informative, though, because it shows not just *whether* there is a difference but also *how big* it plausibly is.\n"
+    "> **CIs and p-values are two views of the same evidence.** A 95% confidence interval that excludes the null value corresponds to a p-value below $\\alpha = 0.05$ (indeed, $95\\%=0.95=1-0.05=1-\\alpha)$). The confidence interval is often more informative, though, because it shows not just *whether* there is a difference but also *how big* it plausibly is.\n"
    ]
   },
   {

notebooks/Notebook_3/Notebook_3b_Statistics_1.py

@@ -283,7 +283,7 @@ def show_widget_or_fallback(widget_fn, fallback_fn):
 #
 # **Back to the penguins.** Suppose our data give a 95% confidence interval of $[750\text{ g},\ 1050\text{ g}]$ for the difference in mean body mass between Gentoo and Adelie penguins. This says that the values between 750 g and 1050 g are all plausible estimates of the true difference.
 #
-# > 💡 **CIs and p-values are two views of the same evidence.** A 95% confidence interval that excludes the null value corresponds to a p-value below $\alpha = 0.05$ (indeed, $95\%=0.95=1-0.05=1-\alpha)$). The confidence interval is often more informative, though, because it shows not just *whether* there is a difference but also *how big* it plausibly is.
+# > **CIs and p-values are two views of the same evidence.** A 95% confidence interval that excludes the null value corresponds to a p-value below $\alpha = 0.05$ (indeed, $95\%=0.95=1-0.05=1-\alpha)$). The confidence interval is often more informative, though, because it shows not just *whether* there is a difference but also *how big* it plausibly is.
 #
 
 # %% [markdown] id="wMvSvNq6Pxpm"

notebooks/Notebook_3/Notebook_3c_Statistics_2.ipynb

@@ -836,15 +836,15 @@
     "id": "ex-diabetes-dataset-inference"
    },
    "source": [
-    "## 🧪 Diabetes Dataset Inference Exercises\n",
+    "## Diabetes Dataset Inference Exercises\n",
     "\n",
     "[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/nerdslab/learningfromdata-course/blob/main/notebooks/Notebook_3/Notebook_3c_Statistics_2.ipynb#scrollTo=ex-diabetes-dataset-inference)\n",
     "\n",
     "You will now apply hypothesis testing and resampling methods to the Diabetes dataset.\n",
     "\n",
     "---\n",
     "\n",
-    "### 🔎 Exercise 1 — Two-Group Comparison (Welch t-test)\n",
+    "### Exercise 1 — Two-Group Comparison (Welch t-test)\n",
     "\n",
     "We split patients into:\n",
     "\n",
@@ -865,7 +865,7 @@
     "\n",
     "---\n",
     "\n",
-    "### 📊 Exercise 2 — ANOVA Across BMI Quartiles\n",
+    "### Exercise 2 — ANOVA Across BMI Quartiles\n",
     "\n",
     "We compared 4 BMI groups using ANOVA.\n",
     "\n",
@@ -880,7 +880,7 @@
     "\n",
     "---\n",
     "\n",
-    "### 🎲 Exercise 3 — Permutation Testing\n",
+    "### Exercise 3 — Permutation Testing\n",
     "\n",
     "Compare:\n",
     "\n",
@@ -895,7 +895,7 @@
     "\n",
     "---\n",
     "\n",
-    "### 🧠 Final Reflection\n",
+    "### Final Reflection\n",
     "\n",
     "Write a short paragraph:\n",
     "\n",

notebooks/Notebook_3/Notebook_3c_Statistics_2.py

@@ -465,15 +465,15 @@ def power_widget(widgets):
 #
 
 # %% [markdown] id="ex-diabetes-dataset-inference"
-# ## 🧪 Diabetes Dataset Inference Exercises
+# ## Diabetes Dataset Inference Exercises
 #
 # [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/nerdslab/learningfromdata-course/blob/main/notebooks/Notebook_3/Notebook_3c_Statistics_2.ipynb#scrollTo=ex-diabetes-dataset-inference)
 #
 # You will now apply hypothesis testing and resampling methods to the Diabetes dataset.
 #
 # ---
 #
-# ### 🔎 Exercise 1 — Two-Group Comparison (Welch t-test)
+# ### Exercise 1 — Two-Group Comparison (Welch t-test)
 #
 # We split patients into:
 #
@@ -494,7 +494,7 @@ def power_widget(widgets):
 #
 # ---
 #
-# ### 📊 Exercise 2 — ANOVA Across BMI Quartiles
+# ### Exercise 2 — ANOVA Across BMI Quartiles
 #
 # We compared 4 BMI groups using ANOVA.
 #
@@ -509,7 +509,7 @@ def power_widget(widgets):
 #
 # ---
 #
-# ### 🎲 Exercise 3 — Permutation Testing
+# ### Exercise 3 — Permutation Testing
 #
 # Compare:
 #
@@ -524,7 +524,7 @@ def power_widget(widgets):
 #
 # ---
 #
-# ### 🧠 Final Reflection
+# ### Final Reflection
 #
 # Write a short paragraph:
 #

notebooks/Notebook_4/Notebook_4b_Classification.ipynb

@@ -1126,7 +1126,7 @@
     "id": "OJmb91Qkhdke"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "\n",
     "> (1) Which $g$'s are \"good\" functions to use for our purposes? What properties do we desire for a \"good\" $g$? What might complicate our problem when trying to solve for $\\beta$?\n",
     "\n",
@@ -2634,7 +2634,7 @@
     "id": "c9sATlc2AOr5"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "\n",
     "Discuss what you can observe from the pairplot for this dataset? Which features appear salient for classification?"
    ]
@@ -4123,7 +4123,7 @@
     "id": "i63GQeRFI898"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "\n",
     "What's happening in the example above? What are two ways that you could try to correct the issue?"
    ]
@@ -4351,7 +4351,7 @@
     "id": "OzSgkBDI0AtG"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "\n",
     "1. Create a set of synthetic experiments to explore how (at least three) different rebalancing approaches work for: (i) different levels of imbalance, (ii) different amounts of noise, and (iii) across different types of data (linear vs nonlinearly separable etc).\n",
     "\n",
@@ -4579,7 +4579,7 @@
     "id": "mnX4ZPTP3RYW"
    },
    "source": [
-    "💥 **Discussion:**\n",
+    "**Discussion:**\n",
     "\n",
     "> Add your discussion here."
    ]

notebooks/Notebook_4/Notebook_4b_Classification.py

@@ -629,7 +629,7 @@ def plot_screening_decision_boundary(model, X, y, ax, title):
 # Where $x_i$ are the inputs or features for every $i^{th}$ data point, and $y_i$ are the corresponding outputs or labels. In general, we can simplify this problem and say that $p=0.5$ and threshold our predictions accordingly.
 
 # %% [markdown] id="OJmb91Qkhdke"
-# 💥 **Challenge:**
+# **Challenge:**
 #
 # > (1) Which $g$'s are "good" functions to use for our purposes? What properties do we desire for a "good" $g$? What might complicate our problem when trying to solve for $\beta$?
 #
@@ -927,7 +927,7 @@ def plot_sigmoid_threshold(threshold=0.5):
 
 
 # %% [markdown] id="c9sATlc2AOr5"
-# 💥 **Challenge:**
+# **Challenge:**
 #
 # Discuss what you can observe from the pairplot for this dataset? Which features appear salient for classification?
 
@@ -1307,7 +1307,7 @@ def train_test_split_with_probablity(X, Y, train_size=250, balanced=True):
 fig.tight_layout()
 
 # %% [markdown] id="i63GQeRFI898"
-# 💥 **Challenge:**
+# **Challenge:**
 #
 # What's happening in the example above? What are two ways that you could try to correct the issue?
 
@@ -1375,7 +1375,7 @@ def train_test_split_with_probablity(X, Y, train_size=250, balanced=True):
 # add your code here
 
 # %% [markdown] id="OzSgkBDI0AtG"
-# 💥 **Challenge:**
+# **Challenge:**
 #
 # 1. Create a set of synthetic experiments to explore how (at least three) different rebalancing approaches work for: (i) different levels of imbalance, (ii) different amounts of noise, and (iii) across different types of data (linear vs nonlinearly separable etc).
 #
@@ -1455,7 +1455,7 @@ def ratio_func(y, multiplier, minority_class):
 
 
 # %% [markdown] id="mnX4ZPTP3RYW"
-# 💥 **Discussion:**
+# **Discussion:**
 #
 # > Add your discussion here.
 

notebooks/Notebook_4/Notebook_4c_PyTorch_Model.ipynb

@@ -340,7 +340,7 @@
     "id": "BkyNqg2DOL6w"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "\n",
     "Write out an expression for the output of the MLP in the figure above. Make clear the dimension of all weight matrices, and assume that a general activation function $\\sigma()$ is used at each hidden layer.\n"
    ]
@@ -486,7 +486,7 @@
     "id": "3IU2V2kvVHst"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "Try some different hyperparameters and test different architectures. Report the accuracies on the train and test sets for at least three different MLP models.\n"
    ]
   },
@@ -2402,7 +2402,7 @@
     "id": "NQaJa_mLVlq0"
    },
    "source": [
-    "💥 **Challenge:**\n",
+    "**Challenge:**\n",
     "Incorporate different types of augmentations into your model, discuss how different augmentations impact the accuracy on the test data.\n",
     "\n",
     ">"

notebooks/Notebook_4/Notebook_4c_PyTorch_Model.py

@@ -132,7 +132,7 @@
 #
 
 # %% [markdown] id="BkyNqg2DOL6w"
-# 💥 **Challenge:**
+# **Challenge:**
 #
 # Write out an expression for the output of the MLP in the figure above. Make clear the dimension of all weight matrices, and assume that a general activation function $\sigma()$ is used at each hidden layer.
 #
@@ -181,7 +181,7 @@
 print("Test accuracy:", clf.score(test_imgs, test_labels))
 
 # %% [markdown] id="3IU2V2kvVHst"
-# 💥 **Challenge:**
+# **Challenge:**
 # Try some different hyperparameters and test different architectures. Report the accuracies on the train and test sets for at least three different MLP models.
 #
 
@@ -919,7 +919,7 @@ def get_model_with_dropout():
 print("Test acc (Early stopping): %.2f" % history["test_acc"][early_stop_epoch])
 
 # %% [markdown] id="NQaJa_mLVlq0"
-# 💥 **Challenge:**
+# **Challenge:**
 # Incorporate different types of augmentations into your model, discuss how different augmentations impact the accuracy on the test data.
 #
 # >