Chapter 4: Recognizing objects
Learning objectives:
Describe object recognition and how feature categorization matters
Distinguish between some of the factors that influence recognition
Provide examples of different feature nets within language
Contrast different theoretical models of recognition for different groups
of objects
Object recognition
Information is processed on multiple different orders
Bottom-up processing
o Data driven
o Taking sensory information and then assembling and integrating
it
(what am i seeing?)
Top-down processing
o Concept driven
o Using models, ideas, and expectations to interpret sensory
information
(is that something I've seen before?)
Integrative agnosia
o Cannot recognize features
Features influencing recognition
Familiarity
o How much you have been exposed to something in the past
Pre and post mask are meant to disrupt the stimulus
High frequency word show up more ( the , and, etc)
Low frequency words words that dont show up as often
Priming
o One input or cue prepares you for an upcoming input or cue
Repetition priming
o Presenting a stimulus more than once
o Processing is more efficient at the second presentation
Word-superiority effect
Ewrlya and lawyer
Context
, o Previously experienced/well-encoded concepts increase
processing rate
Word-superiority effect
o Letters are more accurately and faster to recognize when in a
real word than a jumble
Well-formedness
Well-formedness
o Patterns of letters that follow conventional word rules
Feature nets
Feature nets
o System for recognizing patterns that involve a network of
detectors
o Bottom layer – features
o Higher layers – larger scale objects and composites
The higher the layer the larger the “bigger picture”
o Information travels bottom-up BOTTOM-UP PROCESSING
o Do the net
How do we activate a detector
o Activation level
Current status of a node or detector
Example:
o How “charged” it is or how much energy is
needed
o Response threshold
The amount of activation needed to trigger a detector
response
Actually activating the neuron
What determines a detector’s starting activation level?
o Recency (“warm up” effect)
Activated more recently= higher activation level
o Frequency (“exercise effect”)
Activated more often= higher activation level
Bigram detectors – helps add that additional layer
o Detectors that respond to a pair of inputs
o In this case, letters
Ambiguous inputs – draw graph
Learning objectives:
Describe object recognition and how feature categorization matters
Distinguish between some of the factors that influence recognition
Provide examples of different feature nets within language
Contrast different theoretical models of recognition for different groups
of objects
Object recognition
Information is processed on multiple different orders
Bottom-up processing
o Data driven
o Taking sensory information and then assembling and integrating
it
(what am i seeing?)
Top-down processing
o Concept driven
o Using models, ideas, and expectations to interpret sensory
information
(is that something I've seen before?)
Integrative agnosia
o Cannot recognize features
Features influencing recognition
Familiarity
o How much you have been exposed to something in the past
Pre and post mask are meant to disrupt the stimulus
High frequency word show up more ( the , and, etc)
Low frequency words words that dont show up as often
Priming
o One input or cue prepares you for an upcoming input or cue
Repetition priming
o Presenting a stimulus more than once
o Processing is more efficient at the second presentation
Word-superiority effect
Ewrlya and lawyer
Context
, o Previously experienced/well-encoded concepts increase
processing rate
Word-superiority effect
o Letters are more accurately and faster to recognize when in a
real word than a jumble
Well-formedness
Well-formedness
o Patterns of letters that follow conventional word rules
Feature nets
Feature nets
o System for recognizing patterns that involve a network of
detectors
o Bottom layer – features
o Higher layers – larger scale objects and composites
The higher the layer the larger the “bigger picture”
o Information travels bottom-up BOTTOM-UP PROCESSING
o Do the net
How do we activate a detector
o Activation level
Current status of a node or detector
Example:
o How “charged” it is or how much energy is
needed
o Response threshold
The amount of activation needed to trigger a detector
response
Actually activating the neuron
What determines a detector’s starting activation level?
o Recency (“warm up” effect)
Activated more recently= higher activation level
o Frequency (“exercise effect”)
Activated more often= higher activation level
Bigram detectors – helps add that additional layer
o Detectors that respond to a pair of inputs
o In this case, letters
Ambiguous inputs – draw graph
