CSCE 470 Lecture 28

« previous | Friday, November 1, 2013 | next »

Quiz Topics

Monday

Clustering

• K-Means
• hierarchical
• strengths and weaknesses of each
• understand algorithms

Classification

• problem definition
• how different from clustering?
• Rocchio (linear)
• KNN (nonlinear)

Recommenders

• Collaborative
• Content-based
• cold-start problem

Cav's Trip

Conference

All the companies were there

All are talking about the things we're talking about:

• K-means
• Clustering
• Classifiers
• Recommenders

And "Making Bank" on it

Breakfast at @twitter: @TheRealCaverlee ate #fruit.

Students who work crazy hard at

• Twitter
• Facebook
• Google

Interfaces and Visualization

We have to use the project that we design and build.

What do people search for? (and how?)

Information Needs:

• (Q/A) Where is Texas A&M Located?
• (Ideas) What are some good ideas for landscaping my client's yard?
• (Text mining) What are some promising untried treatments for a disease?

Three types of queries

1. informational
2. navigational
3. transactional

Current methods kinda suck at all of this.

Standard Model

Assumptions:

• Maximize precision and recall simultaneously
• Information need remains static
• Value is in resulting document set.

Information Need:
    Query
    Send to system
    Get results
    Evaluate Results
    if not done
        goto Information Need
    Stop