Rethinking eliminative connectionism. Marcus, G. F Cognit Psychol, 37(3):243–82, 1998.
abstract   bibtex   
Humans routinely generalize universal relationships to unfamiliar instances. If we are told "if glork then frum," and "glork," we can infer "frum"; any name that serves as the subject of a sentence can appear as the object of a sentence. These universals are pervasive in language and reasoning. One account of how they are generalized holds that humans possess mechanisms that manipulate symbols and variables; an alternative account holds that symbol-manipulation can be eliminated from scientific theories in favor of descriptions couched in terms of networks of interconnected nodes. Can these "eliminative" connectionist models offer a genuine alternative? This article shows that eliminative connectionist models cannot account for how we extend universals to arbitrary items. The argument runs as follows. First, if these models, as currently conceived, were to extend universals to arbitrary instances, they would have to generalize outside the space of training examples. Next, it is shown that the class of eliminative connectionist models that is currently popular cannot learn to extend universals outside the training space. This limitation might be avoided through the use of an architecture that implements symbol manipulation.
@Article{Marcus1998a,
  author   = {Marcus, Gary F},
  journal  = {Cognit Psychol},
  title    = {Rethinking eliminative connectionism},
  year     = {1998},
  number   = {3},
  pages    = {243--82},
  volume   = {37},
  abstract = {Humans routinely generalize universal relationships to unfamiliar
	instances. If we are told "if glork then frum," and "glork," we can
	infer "frum"; any name that serves as the subject of a sentence can
	appear as the object of a sentence. These universals are pervasive
	in language and reasoning. One account of how they are generalized
	holds that humans possess mechanisms that manipulate symbols and
	variables; an alternative account holds that symbol-manipulation
	can be eliminated from scientific theories in favor of descriptions
	couched in terms of networks of interconnected nodes. Can these "eliminative"
	connectionist models offer a genuine alternative? This article shows
	that eliminative connectionist models cannot account for how we extend
	universals to arbitrary items. The argument runs as follows. First,
	if these models, as currently conceived, were to extend universals
	to arbitrary instances, they would have to generalize outside the
	space of training examples. Next, it is shown that the class of eliminative
	connectionist models that is currently popular cannot learn to extend
	universals outside the training space. This limitation might be avoided
	through the use of an architecture that implements symbol manipulation.},
  keywords = {Computing Methodologies, Human, Language, Learning, Mental Processes, Models, Theoretical, Stochastic Processes, Support, U.S. Gov't, Non-P.H.S., Cognition, Linguistics, Neural Networks (Computer), Practice (Psychology), Non-U.S. Gov't, Memory, Psychological, Task Performance and Analysis, Time Factors, Visual Perception, Adult, Attention, Discrimination Learning, Female, Male, Short-Term, Mental Recall, Orientation, Pattern Recognition, Visual, Perceptual Masking, Reading, Concept Formation, Form Perception, Animals, Corpus Striatum, Shrews, P.H.S., Visual Cortex, Visual Pathways, Acoustic Stimulation, Auditory Cortex, Auditory Perception, Cochlea, Ear, Gerbillinae, Glycine, Hearing, Neurons, Space Perception, Strychnine, Adolescent, Decision Making, Reaction Time, Astrocytoma, Brain Mapping, Brain Neoplasms, Cerebral Cortex, Electric Stimulation, Electrophysiology, Epilepsy, Temporal Lobe, Evoked Potentials, Frontal Lobe, Noise, Parietal Lobe, Scalp, Child, Language Development, Psycholinguistics, Brain, Perception, Speech, Vocalization, Animal, Discrimination (Psychology), Hippocampus, Rats, Calcium, Chelating Agents, Excitatory Postsynaptic Potentials, Glutamic Acid, Guanosine Diphosphate, In Vitro, Neuronal Plasticity, Pyramidal Cells, Receptors, AMPA, Metabotropic Glutamate, N-Methyl-D-Aspartate, Somatosensory Cortex, Synapses, Synaptic Transmission, Thionucleotides, Action Potentials, Calcium Channels, L-Type, Electric Conductivity, Entorhinal Cortex, Neurological, Long-Evans, Infant, Mathematics, Statistics, Probability Learning, Problem Solving, Psychophysics, Association Learning, Child Psychology, Habituation (Psychophysiology), Probability Theory, Analysis of Variance, Semantics, Symbolism, Behavior, Eye Movements, Macaca mulatta, Prefrontal Cortex, Cats, Dogs, Haplorhini, Photic Stimulation, Electroencephalography, Nervous System Physiology, Darkness, Grasshoppers, Light, Membrane Potentials, Neural Inhibition, Afferent, Picrotoxin, Vision, Deoxyglucose, Injections, Microspheres, Neural Pathways, Rhodamines, Choice Behavior, Speech Perception, Verbal Learning, Dominance, Cerebral, Fixation, Ocular, Language Tests, Random Allocation, Comparative Study, Saguinus, Sound Spectrography, Species Specificity, Audiometry, Auditory Threshold, Calibration, Data Interpretation, Statistical, Anesthesia, General, Electrodes, Implanted, Pitch Perception, Sound Localization, Paired-Associate Learning, Serial Learning, Auditory, Age Factors, Motion Perception, Brain Injuries, Computer Simulation, Blindness, Psychomotor Performance, Color Perception, Signal Detection (Psychology), Judgment, ROC Curve, Regression Analysis, Music, Probability, Arm, Cerebrovascular Disorders, Hemiplegia, Movement, Muscle, Skeletal, Myoclonus, Robotics, Magnetoencephalography, Phonetics, Software, Speech Production Measurement, Epilepsies, Partial, Laterality, Stereotaxic Techniques, Germany, Speech Acoustics, Verbal Behavior, 9892549},
}
Downloads: 0
About
Documentation
Keywords
Search
Users
Terms and Conditions of Use
Privacy Policy
Sitemap
© BibBase.org 2021