Neural Nets - Intro to Cognitive Science: Cognitive Modeling - Fall 2009

Contents

Slides

Overview of Today's Class [1/6, 2009/10/13]

• P2 postmortem
• Neural nets
• Looking ahead to Tuesday
• Looking ahead to Thursday

P2 postmortem [2/6, 2009/10/13]

• Any surprises?
• What did you learn?

Neural nets [3/6, 2009/10/13]

Article

Other links

• Online course questions with links.
• Nervous system (awesome pictures)
• A Neural Nets Tutorial
• A demo program

My key concepts:

• Human
• Neuron
• Axon
• Excitation
• Inhibition
• Recovery
• Artificial
• Nodes
• Links
• Weights
• Layers
• Hidden layer
• Training set
• Backpropagation
• Other learning?

Example

For next Tuesday [4/6, 2009/10/13]

Project 3 description and groups

"Readings"

No written assignment, but there will be a short quiz. Questions like:

• How many input and output units are in the mirror symmetry example?
• How do the (learned) weights on the links in the mirror symmetry example ensure that it will be able to distinguish symmetric and non-symmetric strings?
• In the family tree example, which are the input units, and what do they represent?
• Which are the output units, and what do they represent?
• What does figure 8.4 show? What type of regularities does it pick up?

Key concepts:

• First example is 6-bit binary number strings. Like 000010 or 011011.
• Mirror symmetry.

Midterm QUIZ [5/6, 2009/10/13]

• Format:
• 5% of final grade
• Mostly short answer questions (1-2 sentences)
• Maybe a couple multiple choice.
• Maybe 1 or 2 longer (3-5 sentences)
• Content:
• What you've read and what we've discussed in class.
• Projects too.
• Emphasis on finding connections between topics.
• Some amount of testing conceptual understanding (declarative knowledge?) is necessary.

Review [6/6, 2009/10/13]

What kind of structure?

A suggestion: "the theory, how it works, how it is tested, similarities and differences to the other theories"