dataexposed

Place for resources related to Data Exposed

---

Project maintained by microsoft Hosted on GitHub Pages — Theme by mattgraham

Data Exposed

Something Old, Something New (http://aka.ms/sosn)

This page contains samples, references and other content from the Data Exposed series “Something Old, Something New”. You can find all the Data Exposed Content here.

Episodes

These are the episode areas, in reverse order, and any links or content we post from the shows.

As always, we’ve got a place to ask questions and chat about what you’ve learned! Follow the terms and condidions of our Code of Conduct, and navigate here to post. Make sure you tag your questions with the episode name and date so we know what you’re asking about.

Episode One: Two Data Scientists Walk into a Bar

Date Aired: January 20, 2021

In this exciting episode, Buck and Anna introduce the show, and explain the deep and dark meaning behind the show’s intriguing title. You’ll learn the maths behind Data Science and Machine Learning, why you might want to learn more about them, and then Anna explains a dead-simple way to tell if she wants more coffee.

Probably.

Something Old:

• General:
  • Core Data Science Concepts for Beginners
  • How Much Math Do You Actually Need For Data Science?
  • The Math Behind Machine Learning
  • Lost Islamic Maths
  • Microsoft Research Book on Math and Machine Learning
  • A Personal Journey into Bayesian Networks
  • 100+ Free Books on Data Science
• Statistics
  • History of Statistics
  • Statistics and machine learning: what’s the difference?
  • Beginner guides to learning Statistics for Data Science
  • Book references for Statistics and Machine Learning
• Algebra
  • 5 Reasons to Learn Linear Algebra for Machine Learning
  • Learn Linear Algebra in Five Hours Today with the Wolfram Language
  • Linear Algebra for Machine Learning
  • A gentle approach to Linear Algebra and Machine Learning
  • A Complete Linear Algebra Tutorial in Python
• Calculus
  • Need a refresher in Calculus? There’s a great resource you can use here
  • Essential Math for Data Science: Integrals And Area Under The Curve
  • Calculus in Machine Learning

Something New:

• Process
  • Team Data Science Process (TDSP)
  • Cross-industry standard process for data mining (CRISP-DM)
• Algorithms
  • Algorithm families for Machine Learning
  • Azure Machine Learning Cheat Sheet for selecting algorithms
• Example
  • Coffee use case for Markov Chains

Episode Two: Reduce, Re-use, Recycle

Date Aired: February 24th, 2021

Our hero Buck goes on a trip (virtually though, since there’s a thing) to the data jungle to detangle the tangled. He also notices that your data is a bit…chunky, and shows you how to go low-calorie with your algorithms. Anna shows up to save the day with something really deep. Deep Learning, that is.

References:

• General:
  • Great overall resource on math and Data Science, with links to training locations
  • T-Test - How to use Python to Perform a Paired Sample T-test (marsja.se)
  • Overview of Dimensionality Reduction in Machine Learning
  • Data Literacy is basic, but extremely important. Learn more about it in this free workshop
• EDA:
  • We are in this phase of the Team Data Science Process
  • General Overview of EDA
  • Central Tendency
  • 5 Number Summary
  • Try EDA on a toy (but real) data set
  • EDA Example in R
  • Using Python for Data Visualization
  • A Very cool new library for EDA
• PCA
  • An overview of PCA
  • A practical example in PCA
  • A step by step walkthrough of PCA
  • Python PCA tutorial
• LDA
  • Topic Modeling and Latent Dirichlet Allocation (LDA) in Python
• GDA
  • Gaussian Discriminant Analysis Overview
• Code samples from episode
  • EDA in R and with Azure SQL Database
  • Anscombe’s Quartet Python notebook
  • PCA Example Python notebook

Episode Three: You Have to Draw the Line Somewhere

Date Aired: March 24th, 2021

Join Buck and Anna as they regress. I mean, explain Regression. Anyway, you’ll learn one of the most widely used tools in Statistics and Data Science, and how the math works for its predictions. There’s spreadheets, code, Notebooks and more in this thrill-packed episode, you won’t want to miss it!

References:

• General:
  • [Encyclopedia of Mathematics on Regression]https://encyclopediaofmath.org/index.php?title=Regression_analysis)
  • Really great explainer of the General Linear Model from Statistics
  • Fun site showing spurious correlations
  • [Regression from a financial perspective](https://www.investopedia.com/terms/l/linearrelationship.asp
  • Another good explainer for Regression
  • Susan Sivek does a fantastic job of explaining errors and residuals here, and has some excellent links on other regression formulas
• Code samples from episode
  • Linear regression in Python
  • Logistic Regression in R Programming

Episode Four: Do Kangaroos Prefer Cake Or Frosting, Generally Speaking?

Date Aired: April 28th, 2021

There are two categories of people: People who put people into two categories, and people who don’t. In this exciting episode of “Something Old, Something New” Buck and Anna will show you how to categorize and classify things using math and Machine Learning. As always, be ready to learn and have fun!

(Oh, and we’ll explain that title. Promise.)

References:

• Taxonomies:
  • Taxonomies Explained
  • Naïve Bayes
  • Discriminant Analysis
  • Support Vector Machines
  • Nearest Neighbors
  • Neural Networks
• Classification Math Techniques:
  • K-Means for Beginners
  • Classification Info
  • The math behind K-Means Clustering
  • StatQuest: K-means clustering video
  • General Clusting Math
• Clasification Examples:
  • Anna’s Notebook showing classification examples
  • Naïve Bayes in Data Mining
• Code samples from episode
  • Naïve Bayes example with Code
  • Logistic Regression in R Programming
  • A Clustering Example in R

Episode Five: That's Really Deep

Date Aired: June 2nd, 2021

Buck has often been accused of not thinking about things deeply enough (especially by his wife) so in this Episode Anna will help him take the math and apply it to a branch of Machine Learning called “Deep Learning”. Don’t worry, we’ll take it slow, and give you lots of examples along the way. You’ll learn about the different types of Deep Learning, the math, and the packages and libraries that make it simple to work with.

Also, like any good Hallmark movie, there’s a dog.

References:

• General References:
  • Deep Learning versus Machine Learning
  • Deep Learning Basics: Introduction and Overview
  • A Longer, but REALLY well-done introduction to Deep Learning
  • Introduction to pre-Perceptrons by McCulloch and Pitts
  • You can watch a Perceptron work at this stie
• Math from the episode:
  • Deep Learning math from Mathworks
  • Deep Learning math from DataScienceCentral
  • The math behind Perceptrons
• Libraries and Packages
  • Understanding Deep Learning Frameworks from NVIDIA
  • Tensorflow overview
  • PyTorch introduction
  • Microsoft’s Cognitive Toolkit
  • Caffe introduction
  • Keras for beginners
  • ONNX
  • Another ONNX overview
• Code samples from episode
  • Code from full course on Deep Learning from Microsoft

Episode Six: Not really Artificial, not really Intelligence

Date Aired: June 23rd, 2021

In this episode, Buck and Anna tackle the difficult topic of Ethics in AI. Artificial Intelligence is becoming more powerful, and more prevalent in our lives – but how is it making the choices it does? What are the biases that it has, and how do they get in the code? We’ll explain the issue clearly, and give you specific steps and guidelines you can follow to create ethical AI systems.

References:

• General References:
  • Ethical AI Site from Microsoft
  • AI is getting smarter every day, but it still can’t match the human mind
  • Adversarial machine learning: The underrated threat of data poisoning
  • The mosaic effect and big data
  • Spotting deepfakes isn’t as easy as you might think
  • AI Can Write Disinformation Now—and Dupe Human Readers
• Math from the episode:
  • Bias in statistics
  • Machine learning and bias