Final Exam: Math Behind ML Algorithms

WHAT YOU WILL LEARN

• 

  Discuss residuals in regression
  

  explore least square error
  

  compute the best fit with partial derivatives
  

  calculate r-squared of a regression model
  

  discuss the normal equation
  

  visualize correlations of features
  

  split train and test data and create computations
  

  perform regression and view the predicted values
  

  introduce gradient descent
  

  explore gradients
  

  standardise and shape data for gradient descent
  

  work through a calculation of an epoch
  

  implement a single epoch
  

  identify correlations for performing logistic regression
  

  calculate an s-curve in logistic regression
  

  set up training and testing data for logistic regression
  

  introduce the classification problem
  

  contrast rule-based and ml-based classifiers
  

  recall the structure of a decision tree
  

  define and understand entropy
  

  define and understand information gain
  

  define and calculate gini impurity
  

  recall characteristics of gini impurity
  

  split decision trees based on gini impurity
  

  decide splits for a rule-based decision tree
  

  define a rule-based decision tree
  

  introduce decision trees for continuous values
  

  train an ml-based decision tree
  

  recall how distance-based models work at a high level and identify the use cases of such models
  

  describe the hamming and cosine distance metrics
• 

  recount how the knn and k-means algorithms use distance metrics to perform ml operations
  

  define and visualize two points in a two-dimensional space using python
  

  calculate the euclidean and manhattan distance between two points using scipy as well as your own function
  

  analyze the data used to implement a classification model using k nearest neighbors
  

  implement a function that classifies a point using the k nearest neighbors algorithm
  

  classify test data points using your own knn classifier and evaluate the model using a variety of metrics
  

  code the individual steps involved in performing a clustering operation using the k-means algorithm
  

  define a function that clusters the points in a dataset using the k-means algorithm and then test it
  

  recognize the place of support vector machines (svms) in the machine learning landscape
  

  outline how svms can be used to classify data, how hyperplanes are defined, and the qualities of an optimum hyperplane
  

  recall the qualities of an optimum hyperplane, outline how scaling works with svm, distinguish soft and hard margins, and recognize when and how to use either margin
  

  recall the techniques that can be applied to classify data that are not linearly separable
  

  apply the gradient descent algorithm to solve for the optimum hyperplane
  

  use scikit-learn to generate blob data that is linearly separable
  

  separate a dataset into training and test sets
  

  load a dataset from a csv file into a pandas dataframe and analyze it in preparation for binary classification
  

  generate a heatmap to visualize the correlations between features in a dataset
  

  build and evaluate an svm classifier and recognize the importance of scaling the inputs to such a model
  

  use boxplots, a pair plot, and a heatmap to analyze a dataset in preparation for training a regression model
  

  recall the architecture and components that make up neural networks
  

  mathematical operation of a neuron
  

  compute the weighted sum of inputs with bias
  

  process data in batches and with multiple layers
  

  illustrate relu, leaky relu, and elu activation functions
  

  illustrate step, sigmoid, and tangent activation functions
  

  recall the characteristics of activation functions
  

  describe how unstable gradients can be mitigated using variants of the relu activation function
  

  create a simple neural network with one neuron for regression
  

  illustrate the impact of learning rate and number of epochs of training
  

  illustrate the classification dataset