List of Figures

• A simple HMM where the only observations are j, k, or l
• A unit
• A typical feedforward neural network
• Recurrent neural network architecture
• Dynamic time warping
• Temporal mismatches
• The relationship between channels, frames and streams.
• An example of a stream from the sign recognition domain with the label come.
• General architecture for training
• General architecture for testing
• An example of a piecewise line approximation to some real data
• Event extraction algorithm
• Parameter space and clustering for Blues and Reds domain.
• Rules for telling blues and reds apart.
• K-Means algorithm
• Parameter space with all classes
• Parameter space with come class only
• An example illustrating how the confidence estimate can be made
• Decision tree produced for come