docs: fix typos (#1)

Found via `codespell -S data`

Author: kianmeng

03_gradient/gradient_descend.livemd

@@ -99,7 +99,7 @@ end
 ```elixir
 alias VegaLite, as: Vl
 
-# Generate a sequence that will be used as `weigth`
+# Generate a sequence that will be used as `weight`
 # From -1 to -4, step 0.01
 weights = Enum.map(-100..400, &(&1 / 100))
 
@@ -149,7 +149,7 @@ defmodule C3.LinearRegressionWithoutBias do
   end
 
   @doc """
-  Returns the derivate of the loss curve
+  Returns the derivative of the loss curve
   """
   def gradient(x, y, weight) do
     predictions = predict(x, weight, 0)
@@ -213,7 +213,7 @@ defmodule C3.LinearRegressionWithBias do
   end
 
   @doc """
-  Returns the derivate of the loss curve
+  Returns the derivative of the loss curve
   """
   def gradient(x, y, weight, bias) do
     predictions = predict(x, weight, bias)

04_hyperspace/multiple_regression.livemd

@@ -112,7 +112,7 @@ defmodule C4.MultipleLinearRegression do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   defn gradient(x, y, weight) do
     # in python:

05_discerning/classifier.livemd

@@ -102,7 +102,7 @@ defmodule C5.Classifier do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   defn gradient(x, y, weight) do
     # in python:

06_real/digit_classifier.livemd

@@ -183,7 +183,7 @@ defmodule C5.Classifier do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   defn gradient(x, y, weight) do
     # in python:

07_final/multiclass_classifier.livemd

@@ -186,7 +186,7 @@ defmodule C7.Classifier do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   defn gradient(x, y, weight) do
     # in python:

07_final/sonar_classifier.livemd

@@ -198,7 +198,7 @@ defmodule C7.Classifier do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   @spec gradient(Nx.Tensor.t(), Nx.Tensor.t(), Nx.Tensor.t()) :: Nx.Tensor.t()
   defn gradient(x, y, weight) do

07_final/sonar_seed_comparison.ex

@@ -183,7 +183,7 @@ defmodule C7.Classifier do
   end
 
   @doc """
-  Returns the derivate of the loss curve.
+  Returns the derivative of the loss curve.
   """
   @spec gradient(Nx.Tensor.t(), Nx.Tensor.t(), Nx.Tensor.t()) :: Nx.Tensor.t()
   defn gradient(x, y, weight) do

11_training/neural_network.livemd

@@ -207,7 +207,7 @@ defmodule C11.Classifier do
 
   Each element in the tensors is now constrained between 0 and 1,
   but the activation functions used for `h` and `y_hat` are
-  differents:
+  different:
   - `sigmoid` for the hidden layer `h`
   - `softmax` for the prediction tensor `y_hat`
 

12_classifiers/12_classifiers_01.livemd

@@ -185,7 +185,7 @@ weight = C12.Perceptron.train(x_train, y_train, x_test, y_test, iterations, lr)
 
 The idea:
 
-* Generate a grid of points and use the min/max values from the inital dataset to compute the boundaries.
+* Generate a grid of points and use the min/max values from the initial dataset to compute the boundaries.
 * Classify each point using the weight computed before with the initial dataset
 * Plot the result highlighting the "decision boundary"
 
@@ -200,7 +200,7 @@ y =
   x_train
   |> Nx.slice_along_axis(2, 1, axis: 1)
 
-# Compute the grid bounderies 
+# Compute the grid boundaries 
 x_min =
   x
   |> Nx.to_flat_list()
@@ -223,7 +223,7 @@ y_max =
 
 padding = 0.05
 
-bounderies = %{
+boundaries = %{
   x_min: x_min - abs(x_min * padding),
   x_max: x_max + abs(x_max * padding),
   y_min: y_min - abs(y_min * padding),

12_classifiers/12_classifiers_02.livemd

@@ -191,7 +191,7 @@ weight = C12.Perceptron.train(x_train, y_train, x_test, y_test, iterations, lr)
 
 The idea:
 
-* Generate a grid of points and use the min/max values from the inital dataset to compute the boundaries.
+* Generate a grid of points and use the min/max values from the initial dataset to compute the boundaries.
 * Classify each point using the weight computed before with the initial dataset
 * Plot the result highlighting the "decision boundary"
 
@@ -206,7 +206,7 @@ y =
   x_train
   |> Nx.slice_along_axis(2, 1, axis: 1)
 
-# Compute the grid bounderies 
+# Compute the grid boundaries 
 x_min =
   x
   |> Nx.to_flat_list()
@@ -229,7 +229,7 @@ y_max =
 
 padding = 0.05
 
-bounderies = %{
+boundaries = %{
   x_min: x_min - abs(x_min * padding),
   x_max: x_max + abs(x_max * padding),
   y_min: y_min - abs(y_min * padding),
@@ -477,12 +477,12 @@ _Same steps used with the perceptron_
 
 The idea:
 
-* Generate a grid of points and use the min/max values from the inital dataset to compute the boundaries.
+* Generate a grid of points and use the min/max values from the initial dataset to compute the boundaries.
 * Classify each point using the weight computed before with the initial dataset
 * Plot the result highlighting the "decision boundary"
 
 ```elixir
-# Get x from the tensor (this time `x` is not pre-pended by the bias column)
+# Get x from the tensor (this time `x` is not prepended by the bias column)
 x =
   x_train
   |> Nx.slice_along_axis(0, 1, axis: 1)
@@ -492,7 +492,7 @@ y =
   x_train
   |> Nx.slice_along_axis(1, 1, axis: 1)
 
-# Compute the grid bounderies 
+# Compute the grid boundaries 
 x_min =
   x
   |> Nx.to_flat_list()
@@ -515,7 +515,7 @@ y_max =
 
 padding = 0.05
 
-bounderies = %{
+boundaries = %{
   x_min: x_min - abs(x_min * padding),
   x_max: x_max + abs(x_max * padding),
   y_min: y_min - abs(y_min * padding),

12_classifiers/circles_data.livemd

@@ -189,7 +189,7 @@ weight = C12.Perceptron.train(x_train, y_train, x_test, y_test, iterations, lr)
 
 The idea:
 
-* Generate a grid of points and use the min/max values from the inital dataset to compute the boundaries.
+* Generate a grid of points and use the min/max values from the initial dataset to compute the boundaries.
 * Classify each point using the weight computed before with the initial dataset
 * Plot the result highlighting the "decision boundary"
 
@@ -204,7 +204,7 @@ y =
   x_train
   |> Nx.slice_along_axis(2, 1, axis: 1)
 
-# Compute the grid bounderies 
+# Compute the grid boundaries 
 x_min =
   x
   |> Nx.to_flat_list()
@@ -227,7 +227,7 @@ y_max =
 
 padding = 0.05
 
-bounderies = %{
+boundaries = %{
   x_min: x_min - abs(x_min * padding),
   x_max: x_max + abs(x_max * padding),
   y_min: y_min - abs(y_min * padding),
@@ -465,12 +465,12 @@ _Same steps used with the perceptron_
 
 The idea:
 
-* Generate a grid of points and use the min/max values from the inital dataset to compute the boundaries.
+* Generate a grid of points and use the min/max values from the initial dataset to compute the boundaries.
 * Classify each point using the weight computed before with the initial dataset
 * Plot the result highlighting the "decision boundary"
 
 ```elixir
-# Get x from the tensor (this time `x` is not pre-pended by the bias column)
+# Get x from the tensor (this time `x` is not prepended by the bias column)
 x =
   x_train
   |> Nx.slice_along_axis(0, 1, axis: 1)
@@ -480,7 +480,7 @@ y =
   x_train
   |> Nx.slice_along_axis(1, 1, axis: 1)
 
-# Compute the grid bounderies 
+# Compute the grid boundaries 
 x_min =
   x
   |> Nx.to_flat_list()
@@ -503,7 +503,7 @@ y_max =
 
 padding = 0.05
 
-bounderies = %{
+boundaries = %{
   x_min: x_min - abs(x_min * padding),
   x_max: x_max + abs(x_max * padding),
   y_min: y_min - abs(y_min * padding),

16_deeper/16_deeper.livemd

@@ -80,7 +80,7 @@ defmodule C16.EchidnaDataset do
 
     # After MinMaxScaling, the distributions are not centered
     # at zero and the standard deviation is not 1.
-    # Thefore, subtract 0.5 to rescale data between -0.5 and 0.5
+    # Therefore, subtract 0.5 to rescale data between -0.5 and 0.5
     (x_raw - min) / (max - min) - 0.5
   end
 
@@ -287,7 +287,7 @@ defmodule C16.Plotter do
     # Get y from the tensor
     y = Nx.slice_along_axis(inputs, 2, 1, axis: 1)
 
-    # Compute the grid bounderies 
+    # Compute the grid boundaries 
     x_min = x |> Nx.to_flat_list() |> Enum.min()
     x_max = x |> Nx.to_flat_list() |> Enum.max()
     y_min = y |> Nx.to_flat_list() |> Enum.min()
@@ -354,7 +354,7 @@ Axon.Display.as_graph(new_model, template)
 # epochs are defined in the previous model's training
 
 # Set `eps` option in the RMSprop to prevent division by zero (NaN)
-# By deafult in Axon is 1.0e-8, I tried with 1.0e-7 (Keras default) and
+# By default in Axon is 1.0e-8, I tried with 1.0e-7 (Keras default) and
 # it was still returning NaN.
 epsilon = 1.0e-4
 

17_overfitting/17_overfitting.livemd

@@ -83,7 +83,7 @@ defmodule C17.EchidnaDataset do
 
     # After MinMaxScaling, the distributions are not centered
     # at zero and the standard deviation is not 1.
-    # Thefore, subtract 0.5 to rescale data between -0.5 and 0.5
+    # Therefore, subtract 0.5 to rescale data between -0.5 and 0.5
     (x_raw - min) / (max - min) - 0.5
   end
 
@@ -132,7 +132,7 @@ validation_data = [{x_validation, y_validation}]
 epochs = 30_000
 
 # Set `eps` option in the RMSprop to prevent division by zero (NaN)
-# By deafult in Axon is 1.0e-8, I tried with 1.0e-7 (Keras default) and
+# By default in Axon is 1.0e-8, I tried with 1.0e-7 (Keras default) and
 # it was still returning NaN.
 epsilon = 1.0e-4
 

19_beyond/beyond_vanilla_networks.livemd

@@ -44,7 +44,7 @@ rows = 4
 
 key = Nx.Random.key(42)
 
-# Compute random indeces
+# Compute random indices
 indices =
   {elem(images_shape, 0) - 1}
   |> Nx.iota()

README.md

@@ -52,7 +52,7 @@ If you want to launch Livebook via Docker:
 
 ### Differences between Livebook an Jupyter books
 
-* I could replicate all the different Jupyter books in Elixir with Livebook/Nx/Axon, apart from the 2nd section of Chapter 17, where the book introduces L1/L2 regularization tecniques and these are not supported by [Axon](https://github.com/elixir-nx/axon) out of the box (more details in the corresponding Livebook).
+* I could replicate all the different Jupyter books in Elixir with Livebook/Nx/Axon, apart from the 2nd section of Chapter 17, where the book introduces L1/L2 regularization techniques and these are not supported by [Axon](https://github.com/elixir-nx/axon) out of the box (more details in the corresponding Livebook).
 
 ### Code Style