DemystifyingEd-Tech Algorithms

Kris Shaffer, Ph.D.Univ. of Mary Washington

kris.shaffermusic.com / @krisshafferview presentation (with live links) at kris.shaffermusic.com/edtechalgorithmsfork presentation at github.com/kshaffer/edtechalgorithms

kris.shaffermusic.com/edtechalgorithms

what constitutes 'adaptive'learning?

is algorithmic pedagogy 'personalized'?

Data science is data science.

source: Hada del lago (Flickr) CC BY-NC-ND

More data beatsa better algorithm.

source: Terry Freedman

machine learning:a crash course

Machine learning involves the use of computers to perform an algorithmic analysis of data in order to discover relationships in that data, often with the goal of making predictions about future data.

classification

categorical data

regression

continuous data

machine learning

Gain a more nuanced understanding of relationships within the data, and make better predictions about future outcomes, than could be done by a human with pencil, paper, and a calculator.

supervised learning

Both data and outcomes are known.

Algorithm searches for relationships.

unsupervised learning

Data is known.

Both outcomes and relationships are emergent.

simple algorithms

Comparing two data parameters of equal length

Correlation

comparing two streams of continuous data

Ex.: Is there a relationship between number of dogs owned and annual vacuum filter purchases?

Chi-Squared Test

comparing two streams of categorical data

Ex.: Is there a relationship between political party membership and religious affiliation?

ANoVA (Analysis of Variance)

comparing a stream of categorical data with a stream of continuous data

Ex.: Is there a relationship between house architecture (ranch, colonial, brownstone, tudor, etc.) and selling price?

supervised regression algorithms

linear regression

What is the relationship between contributing factor(s) and a continuous data result?

Ex.: What is the impact of home size and age on sale price?

linear regression

source: Playing with machine learning: Linear Regression

logistic regression

What is the relationship between contributing factor(s) and a categorical result?

Ex.: Do calorie intake and fat intake contribute to heart attacks?

supervised classification algorithms

How do specific features influence classification? How can that information be used to classify new observations?

Ex.: Voice recognition, OCR, face detection.

K-nearest neighbor

source: Classification of Hand-written Digits

decision tree

source: Wikipedia

random forest

source: Random forest classifier combined with feature selection for breast cancer diagnosis and prognostic

neural networks

source: Extreme Tech

unsupervised classification algorithms

Given a known set of features, what kinds of classifications emerge from the data?

Ex.: Recommendation systems.

topic modeling

source: Topic Modeling and Network Analysis

cluster analysis

source: Comparing Players Using Cluster Analysis

sequential modeling

Given an event or sequence of events, what is the likelihood that a particular event will happen next?

markov models

source: Stack Overflow

exploration

Knewton Adaptive Learning: Building the world’s most powerful education recommendation engine

Mark anything that sounds like statistics, machine learning, or algorithms.

• Is it classification or regression?
• Is it supervised or unsupervised?
• What are the data sources/features?
• What are the predicted outputs?
• What is the pedagogical intention?

going deeper

What are some algorithms or data analysis examples from your own educational practice? (or your institutions?)

What about your current pedagogy is (quasi-)algorithmic? What is the pedagogical intention?

DemystifyingEd-Tech Algorithms

Kris Shaffer, Ph.D.Univ. of Mary Washington

kris.shaffermusic.com / @krisshafferview presentation (with live links) at kris.shaffermusic.com/edtechalgorithmsfork presentation at github.com/kshaffer/edtechalgorithms