Intro to Machine Learning 

Nanodegree Program Syllabus  
 
 
In collaboration with:  
 
Level:​ Intermediate 
Duration:​ 3 months (10 hours/week) 
Before You Start  
Educational  Objectives:  After  this  program,  students  will  be  able  to  use Python and SQL to access and analyze 
data  from  several  different data sources, build predictive models using a variety of unsupervised and supervised 
machine  learning  techniques  and  perform  feature engineering to improve the performance of machine learning 
models.  They  will  also  be  able  to  optimize,  tune,  and  improve  algorithms  according  to  specific  metrics  like 
accuracy and speed and compare the performances of learned models using suitable metrics. 

Length  of  Program:  The  program  is  delivered  in  1  term  spread  over  4  months.  On  average,  students will need 
to  spend  about  10  hours  per  week  in  order  to  complete  all  required  coursework,  including  lecture  and  project 
time.  

Prerequisites: To optimize your chances of success in this program, we recommend having intermediate Python 
programming  knowledge  with  at  least  40hrs  of  programming  experience,  familiarity  with  data  structures  and 
experience  with  libraries.  Students  should  also  have  a  basic  knowledge  of  probability  and  statistics  including 
calculating  the  probability  of  an  event  and  familiarity  with  terms  like  the  mean  and  variance  of  a  probability 
distribution. 

The  following  courses  can  help  you  prepare  for  this  program:  ​AI  Programming  with  Python  Nanodegree 
Program​, ​Intro to Programming Nanodegree Program. 

version 1.0  
1 
Nanodegree Program Overview Page:​ c​ lick here  

Nanodegree Program Info  

The  Intro  to  Machine  Learning  Nanodegree  program is to help students learn machine learning techniques such 
as  data  transformation  and  algorithms  that  can  find  patterns  in  data  and  apply  machine  learning algorithms to 
tasks  of  their  own design. This Nanodegree ​is intended for students with experience in Python, who have not yet 
studied Machine Learning topics.

 
Module 1: Supervised Learning 
In  this  lesson,  you  will  learn  about  supervised  learning,  a  common  class  of  methods  for  model  construction.  In 
Supervised  Learning,  algorithms  learn  from  labeled  data.  The algorithm determines which label should be given 
to new data based on pattern and associating the patterns to the unlabeled new data. 
 
Project 1: Find Donors for Charity 
CharityML  is  a  fictitious  charity  organization  located  in the heart of Silicon Valley that was established to provide 
financial  support  for  people  eager  to  learn  machine  learning.  To  expand  their  potential  donor  base,  CharityML 
has  decided  to  send  letters  to  residents  of  California, but to only those most likely to donate to the charity. Your 
goal  will  be  to  evaluate  and  optimize  several  different  supervised  learning  algorithms  to  determine  which 
algorithm will provide the highest donation yield while under some marketing constraints.

Lesson Title  Learning Outcomes 

REGRESSION  Learn the difference between Regression and Classification and train a 
Linear Regression model to predict values. Learn to predict states using 
Logistic Regression. 

PERCEPTION ALGORITHMS  In  this  lesson  you will learn the definition of a perceptron as a building block 

for neural networks, and the perceptron algorithm for classification. 

DECISION TREES  Train  Decision  Trees  to  predict  states  and  use  Entropy  to  build  decision 
trees, recursively. 

NAIVE BAYES’  Learn  Bayes’  rule,  and  apply  it  to  predict  cases of spam messages using the 
Naive  Bayes  algorithm  and  train  models  using Bayesian Learning. Complete 
an exercise that uses Bayesian Learning for natural language processing. 

SUPPORT VECTOR MACHINES  Learn  to  train  a  Support  Vector  Machines  to separate data, linearly and use 
Kernel Methods in order to train SVMs on data that is not linearly separable. 

ENSEMBLE OF LEARNERS  Enhance  traditional  algorithms  via  boosting  and  learn  and  apply  Random 
Forest  algorithms.  Use  AdaBoost  and  evaluate  the  performance  of boosted 
models 

version 1.0  
2 
 EVALUATION METRICS  Discover  different  metrics  to  measure  model  success  and  calculate 
accuracy, precision, and recall to measure the performance of your models. 

TRAINING AND TUNING MODELS  Train  and  test  models  with  Scikit-learn,  then  choose  the  best  model  using 
evaluation techniques like cross-validation and grid search. 
 
Module 2: Neural Networks 
In this lesson, you’ll learn the foundations of neural network design and training in PyTorch.  
Project 2: Create an Image Classifier 
Implementing an image classification application using a deep neural network. This application will train a deep 
learning model on a dataset of images. It will then use the trained model to classify new images. You will develop 
your code in a Jupyter notebook to ensure your implementation works well. 

Lesson Title  Learning Outcomes 

INTRODUCTION TO NEURAL  Learn the foundations of deep learning and neural networks and 
NETWORKS  implement gradient descent and backpropagation in Python 

IMPLEMENTING GRADIENT  Implement gradient descent using NumPy matrix multiplication. 

DESCENT 

TRAINING NEURAL NETWORKS  In this lesson you will discover several techniques to effectively train a 
neural network and prevent overfitting of training data and learn best 
practices for minimizing the error of a network. 

DEEP LEARNING WITH PYTORCH  Learn how to use PyTorch for building deep learning models. 

 
 
Module 3: Unsupervised Learning 
In this lesson, you’ll learn to implement unsupervised learning methods for different kinds of problem domains. 
Unsupervised learning uses data that is neither classified nor labeled and allows the algorithm to act on that 
information without guidance. 
Project 3: Creating Customer Segments 
In this project, you will apply unsupervised learning techniques on product spending data collected for 
customers of a wholesale distributor to identify customer segments hidden in the data. You will first explore and 
pre-process the data by scaling each product category and then identifying (and removing) unwanted outliers. 
With the cleaned customer spending data, you will apply PCA transformations to the data and implement 
clustering algorithms to segment the transformed customer data. Finally, you will compare the segmentation 
found with an additional labeling and consider ways this information could assist the wholesale distributor with 
future service changes. 

Lesson Title  Learning Outcomes 

version 1.0  
3 
 CLUSTERING  Learn the basics of clustering data and cluster data with the K-means 
algorithm. 

HIERARCHICAL AND  Cluster data with Single Linkage Clustering and with DBSCAN, a clustering 
DENSITY-BASED CLUSTERING  method that captures the insight that clusters are dense group of points. 

GAUSSIAN MIXTURE MODELS  Cluster data with Gaussian Mixture Models and optimize Gaussian 
Mixture Models with an Expectation Maximization. 

DIMENSIONALITY REDUCTION  Learn to reduce the dimensionality of the data using Principal 
Component Analysis and Independent Component Analysis. 
 
 
 

Contact Info  
While going through the program, if you have questions about anything, you can reach us at​ ​enterprise​[email protected]​.

 
 
 
 

version 1.0  
4