Psy+261+Week+4

Sensori-Motor Coordination
 * Neural activations sent from the brain throughout the body's muscles control movement.
 * The brain figures out how to move by vector coding. This sends a pattern of activation levels to all muscles at the same time.
 * Feedforward networks allow us to transform any incoming sensory vectors into an appropriate outgoing motor vector, into a specific motor vector.
 * Feedforward networks cannot represent time.
 * Intelligence handles the coordination of behavior to go with perception.

Recurrencies and Time
 * A real brain guiding a real body must generate, not a single motor vector for one-time delivery to its many muscles, but an ongoing sequence of motor vectors whose changes over time will produce the right bodily changes over time.
 * The nervous system never rests since there are neural activations sent to motor vectors at all times.

Recurrencies and Concepts
 * Ambiguous figures can be seen in different ways.
 * A trained network has a strong tendency to “complete” input vectors that are slightly degraded
 * Biasing the vectorial activity at layer 2 of the network may lead to the production of prototype vectors that are flatly inappropriate to the external reality.
 * When an image is seen, as in when you see the Dalmatian above, it is difficult if not impossible to see the initial confusion that you first saw.
 * Recurrent networks display a feature of being predictable only within statistical limits.
 * The vector to vector transformations within a recurrent network are nonlinear and do not follow a straight line.

Study Questions and Powerpoint Slides
 * Sensori-Motor Coordination Slides [[file:Feb8.ppt]]
 * Recurrencies Slides
 * Chapter 4 Study Questions [[file:Churchland Chapt 4.doc]]
 * Chapter 5 Study Questions[[file:Churchland Chapt 5.doc]]