patterns Flashcards
template matching
We match objects to stored representations in our head (identification of patterns through comparing the pattern of neural excitation arising from the retina with patterns to templates stored in our memory)
template matching specifics
- Direct one-to-one match, wholistic (what is stored in our memory is matched to outside information to make a connection)
- Works well for pattern recognition machines, but there are too many limitations to work for human pattern recognition (ex. Slight deviations in shape, size, orientation, etc.)
EX. recognizing A and R despite their similar compositions
normalization
- If all pattern recognition depended on normalization & TMing processes, a tremendous amount of normalization processes would be needed
(ex. Looking at trees without leaves might require the normalization/transformation process of putting leaves on the tree to match with our prototypical idea of a tree)
(ex. New info is normalized into something more recognizable, then reflected onto the prototype → I see a chihuahua → I normalize it to my prototype of a dog → I see that it’s a dog) - Prototypes get stored and become the goal of normalization
feature detection vs. feature analysis
Both assume that elementary features are used to identify patterns, but they differ in assumptions of how these features are processed
feature detection
- Assumes that the perceptual system detects the presence or absence of specific features and uses this info to determine what stimulus the person is perceiving
- Present or not present, no in between
EX. Smudge on the “R” obscuring the curved part → “R” isn’t there because it’s not whole
feature analysis
- The features are not necessarily assumed to be either definitely present or definitely absent
- Strong evidence or weak evidence
EX. Smudge on the “R” obscuring the curved part → Enough evidence is there to match stored features to seen features of the letter “R” to conclude that the letter is there
EX. Pandemonium model
feature comparison
- We store features, we take them apart, and we match them
- Many common elements are shared by different stimuli
- More cognitive economy because less information is stored (only recognizing certain shared features, EX. Letter elemental features like p, b, q, and d)
- Human pattern recognition is very flexible and pattern recognition must account for this flexibility
bottom-up model
Stimulus-driven = Bottom-up (taking the stimulus and recognizing it by the stimulus)
top-down model
Context-driven = Top-down (ex. I know Prof. Binder likes coffee, so I can recognize the object in her hand as a mug)
advantages of feature comparison
1) It is possible to identify the relationships among features
that are critical for identifying a given letter (A = / – )
2) The models can more easily account for variability in the
stimuli
3) Less storage required
4) The processes by which stimuli are compared to information stored in memory are also quite different in the two theories.
- wholistic process (template)
- analytic process (feature comparison)
visual search tasks (support for FC theories)
- Z is embedded in distractors → Either share many straight-line features with the target (X, W, V) or curved features (O, C)
- Participants detected target much faster when the distractors had features dissimilar to the target → Pandemonium model
support for FC theories
- More shared features = More likely to be confused (pandemonium again)
EX. B vs. 13 = CONTEXT!!!!
EX. Dick vs. Duck = CONTEXT!!!!
- Paradox because usually things should be more easily remembered when the thing is simpler, but words are remembered more often
word superiority effect
More likely to place the flashed letter in the right place when an actual world is flashed → Easier to identify letters when presented in a word than when presented individually or jumbled
Interactive Activation Model of Word Recognition
connectionist model
1) Processing at 3 different levels (feature, letter, word)
2) Perceptual processes occur in parallel within these different levels
3) The model is an interactive activation model that includes both bottom-up and top-down processes (ex. Knowledge of English words, information about features, etc.)
macrostructure
Macro specifies how the levels of the model are organized and how information is passed between levels
microstructure
Micro specifies in detail how the information is processed by the modes within the system
