Although it is commonly assumed that rule-based models generalize more effectively than do connec... more Although it is commonly assumed that rule-based models generalize more effectively than do connectionist models, the comparison is often confounded by pitting handwritten rules against learned connections. Three case studies from cognitive development show that, under the constraint that both types of models learn their representations from equivalent examples, generalization is consistently superior in connectionist models.
Autonomous learning is the ability to learn effectively without much external assistance, which i... more Autonomous learning is the ability to learn effectively without much external assistance, which is a desirable characteristic in both engineering and computational-modeling. We extend a constructive neural-learning algorithm, sibling-descendant cascade-correlation, to monitor lack of progress in learning in order to autonomously abandon unproductive learning. The extended algorithm simulates results of recent experiments with infants who abandon learning on difficult tasks. It also avoids network overtraining effects in a more realistic manner than conventional use of validation test sets. Some contributions and limitations of constructive neural networks for achieving autonomy in learning are briefly assessed.
We propose a constructive neural-network model comprised of deterministic units which estimates a... more We propose a constructive neural-network model comprised of deterministic units which estimates and represents probability distributions from observable events -- a phenomenon related to the concept of probability matching. We use a form of operant learning, where the underlying probabilities are learned from positive and negative reinforcements of the inputs. Our model is psychologically plausible because, similar to humans, it learns to represent probabilities without receiving any representation of them from the external world, but rather by experiencing individual events. Also, we discuss how the estimated probabilities can be used in a setting with deterministic units to produce matching behaviour in choice. Our work is a step towards understanding the neural mechanisms underlying probability matching behavior by specifying processes at the algorithmic level.
Although the reasons for the success of computer simulations of psychology are often difficult to... more Although the reasons for the success of computer simulations of psychology are often difficult to identify, it is possible to make some progress through systematic experimentation. Reasons for the success of cascadecorrelation models of cognitive development are identified in two case studies. Cascade-correlation is a generative neural network model that constructs its own topology as it recruits hidden units. Capturing correct stage sequences in the integration of velocity, time, and distance cues requires a system that grows in computational power while it learns. Static networks are either too weak or too powerful to capture the full range of stages. Simulating the variation and stages in acquisition of the semantics of English personal pronouns requires sensitivity to differing amounts of addressed and non-addressed speech. Just as with children, networks benefit from the opportunity to hear personal pronouns used in exchanges between o ther speakers. Other simulations suggest t...
Learning by imitation is a powerful process for acquiring new knowledge, but there has been littl... more Learning by imitation is a powerful process for acquiring new knowledge, but there has been little research exploring imitation’s potential in the problem solving domain. Classical problem solving techniques tend to center around reinforcement learning, which requires significant trial-and-error learning to reach successful goals and problem solutions. Heuristics, hints, and reasoning by analogy have been favored as improvements over reinforcement learning, whereas imitation learning has been regarded as rote memorizing. However, research on imitation learning in animals and infants suggests that what is being learned is the overall arrangement of actions (sequencing and planning). Applied to problem solving, this suggests that imitation learning might enable a problem solver to infer a complex hierarchical problem representation from observation alone. We compared three types of learning in problem solving tasks: imitation learning (a group that viewed successful problem solving de...
Learning the Structure of Abstract Groups Dirk Schlimm ([email protected]) Department of Phi... more Learning the Structure of Abstract Groups Dirk Schlimm ([email protected]) Department of Philosophy and School of Computer Science, McGill University 855 Sherbrooke St. W., Montreal, QC H3A 2T7, Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University 1205 Penfield Avenue, Montreal, QC H3A 1B1, Canada Abstract It has recently been shown that neural networks can learn particular mathematical groups, for example, the Klein 4- group (Jamrozik & Shultz, 2007). However, there are groups with any number of elements, all of which are said to instantiate the abstract group structure. Learning to differentiate groups from other structures that are not groups is a very difficult task. Contrary to some views, we show that neural networks can learn to recognize finite groups consisting of up to 4 elements. We present this problem as a case study that exhibits the advantages of knowledge-based learning over knowledge-free l...
The Cascade-Correlation connectionist architecture was used to model human cognitive development ... more The Cascade-Correlation connectionist architecture was used to model human cognitive development on balance scale problems. The simulations were characterized by gradual expansion of the training patterns, training bias in favor of equal distance problems, and test problems balanced for torque difference. Both orderly rule stages and torque difference effects were obtained. Analyses of the development of network structure revealed progressive sensitivity to distance information. It was noted that information salience effects, such as that for torque difference, are particularly difficult to capture in symbolic level models.
A knowledge-based constructive learning algorithm, KBCC, simplifies and accelerates the learning ... more A knowledge-based constructive learning algorithm, KBCC, simplifies and accelerates the learning of parity and chessboard problems. Previously learned knowledge of simpler versions of these problems is recruited in the service of learning more complex versions. A learned solution can be viewed as a composition in which the components are not altered, showing that concatenative compositionality can be achieved in neural terms. The idea of compositionality is that mental representations are built out of parts and possess a meaning that is derived from the meanings of the parts and how the parts are combined. A symbolic expression can exhibit what is called concatenative compositionality, which means that the expression incorporates its constituents without changing them. This kind of compositionality makes it possible for symbolic propositions to express the hierarchical, tree-like structure of sentences in a natural language. It has been argued that compositionality exists in a wide ...
Neural-network Modelling of Bayesian Learning and Inference Milad Kharratzadeh (milad.kharratzade... more Neural-network Modelling of Bayesian Learning and Inference Milad Kharratzadeh ([email protected]) Department of Electrical and Computer Engineering, McGill University, 3480 University Street Montreal, QC H3A 2A7 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract We propose a modular neural-network structure for imple- menting the Bayesian framework for learning and inference. Our design has three main components, two for computing the priors and likelihoods based on observations and one for apply- ing Bayes’ rule. Through comprehensive simulations we show that our proposed model succeeds in implementing Bayesian learning and inference. We also provide a novel explanation of base-rate neglect, the most well-documented deviation from Bayes’ rule, by modelling it as a weight decay mechanism which increases entropy. Keywords: Neural-networ...
Contrary to a recent claim that neural network models are unable to account for data on infant ha... more Contrary to a recent claim that neural network models are unable to account for data on infant habituation to artificial language sentences, the present simulations show successful coverage with cascade-correlation networks using analog encoding. The results demonstrate that a symbolic rule-based account is not required by the infant data. One of the fundamental issues of cognitive science continues to revolve around which type of theoretical model better accounts for human cognition -- a symbolic rulebased account or a sub-symbolic neural network account. A recent study of infant habituation to expressions in an artificial language claims to have struck a damaging blow to the neural network approach (Marcus, Vijayan, Rao, & Vishton, 1999). The results of their study show that 7month-old infants attend longer to sentences with unfamiliar structures than to sentences with familiar structures. Because of certain features of their experimental design and their own unsuccessful neural n...
The non-linear complexities of neural networks make network solutions difficult to understand. Sa... more The non-linear complexities of neural networks make network solutions difficult to understand. Sanger's contribution analysis is here extended to the analysis of networks automatically generated by the cascade-correlation learning algorithm. Because such networks have cross connections that supersede hidden layers, standard analyses of hidden unit activation patterns are insufficient. A contribution is defined as the product of an output weight and the associated activation on the sending unit, whether that sending unit is an input or a hidden unit, multiplied by the sign of the output target for the current input pattern. Intercorrelations among contributions, as gleaned from the matrix of contributions x input patterns, can be subjected to principal components analysis (PCA) to extract the main features of variation in the contributions. Such an analysis is applied to three problems, continuous XOR, arithmetic comparison, and distinguishing between two interlocking spirals. In...
Development of Prototype Abstraction and Exemplar Memorization Irina Baetu ([email protected]... more Development of Prototype Abstraction and Exemplar Memorization Irina Baetu ([email protected]) Department of Psychology, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada likely to rely on prototypes at the beginning of a categorization task, and as training progresses they rely more on memorized exemplars (Horst, Oakes, & Madole, 2005; Minda & Smith, 2001; Smith & Minda, 1998). These studies are consistent with a shift from early prototype use to later exemplar memorization. In addition to the amount of experience with a categorization task, category structure also influences which type of information is most used. Better-structured categories can be represented as separate clusters in psychological space, whereas poorly structured categories overlap with each other (Figure 1). Smith and Mi...
Strategies, Heuristics and Biases in Complex Problem Solving Frederic Dandurand ([email protected]... more Strategies, Heuristics and Biases in Complex Problem Solving Frederic Dandurand ([email protected]) Department of Psychology, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada Kristine H. Onishi ([email protected]) Department of Psychology, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada major blocks or sets during preliminary analyses that may limit participants’ ability to explore promising areas of the problem space: a symmetry bias, a simplicity bias, and a tendency to ignore some important aspect of the task. We used a complex puzzle involving the weighing of balls. Participants had to find, with three uses of a scale, the one ball that was either heavier or lighter than the rest of a set of 12 balls. Any combination of balls was a...
Learning the Structure of a Mathematical Group Anna Jamrozik ([email protected]) Depar... more Learning the Structure of a Mathematical Group Anna Jamrozik ([email protected]) Department of Psychology, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract The Klein 4-group A mathematical group is a set of operations that satisfies the properties of identity, inverse, associativity, and closure. Understanding of group structure can be assessed by changing elements and operations over different versions of the same underlying group. Participants learned this structure more quickly over four different, successive versions of a subset of the Klein 4-group, suggesting some understanding of the group structure (Halford, Bain, Mayberry, & Andrews, 1998). Because an artificial neural network learning the task failed to improve, it was argued that such models are incapable of learning a...
Many of the problems of children's cognitive development have been difficult and longstanding... more Many of the problems of children's cognitive development have been difficult and longstanding. Among these are issues of how children represent knowledge, whether stages exist, and how children achieve transitions between stages. Despite a century of scientific evidence on child development, comprehensive theoretical understanding of these issues has remained elusive. Part of the reason is that the problems of psychological development are too complex for traditional verbal theories of development. Nonetheless, it has become clear in recent years that considerable leverage on these problems can be gained by applying computational modeling, particularly using artificial neural networks. This is because computational modeling is a good way to capture complex processes, neural networks capture developmental phenomena in a natural way, and successful models can be examined in detail to discover insights into the phenomena that they simulate (Shultz, 2001).
Recent experiments reveal that 6- to 12-month-old infants can learn probabilities and reason with... more Recent experiments reveal that 6- to 12-month-old infants can learn probabilities and reason with them. In this work, we present a novel computational system called Neural Probability Learner and Sampler (NPLS) that learns and reasons with probabilities, providing a computationally sufficient mechanism to explain infant probabilistic learning and inference. In 24 computer simulations, NPLS shows how probability distributions can emerge naturally from neural-network learning of event sequences, providing a novel explanation of infant probabilistic learning and reasoning. Three mathematical proofs show how and why NPLS simulates the infant results so accurately. The results are situated in relation to seven other active research lines. This work provides an effective way to integrate Bayesian and neural-network approaches to cognition. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
Assessing Generalization in Connectionist and Rule-based Models Under the Learning Constraint Tho... more Assessing Generalization in Connectionist and Rule-based Models Under the Learning Constraint Thomas R. Shultz ([email protected]) Department of Psychology; McGill University Montreal, QC H3C 1B1 Canada Abstract Although it is commonly assumed that rule-based models generalize more effectively than do connectionist models, the comparison is often confounded by pitting hand-written rules against learned connections. Three case studies from cognitive development show that, under the constraint that both types of models learn their representations from equivalent examples, generalization is consistently superior in connectionist models. Generalization Problems A significant part of the ongoing debate between rule- based and connectionist modeling in psychology has focused on the ability to generalize. A common claim from supporters of the classical, symbolic approach is that rule-based models are superior because they generalize more effectively than do connectionist models (Ling ...
A Fresh Look at Vocabulary Spurts Frederic Dandurand ([email protected]) Department of... more A Fresh Look at Vocabulary Spurts Frederic Dandurand ([email protected]) Department of Psychology, Universite de Montreal, 90 ave. Vincent-d'Indy Montreal, QC H2V 2S9 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract There is currently rather little agreement about the existence of, and explanation for, a vocabulary spurt in children during the second year. Here we apply a Functional Data Analysis- based technique called Automatic Maxima Detection to the problem of finding vocabulary spurts in a sample of 20 children. Even with considerable smoothing of the data, children were found to exhibit multiple vocabulary spurts of varying intensity and location. These results should provide a clearer target for researchers interested in detecting and explaining these deviations from linear growth. Keywords: vocabulary spurts; functional data an...
Culture is considered an evolutionary adaptation that enhances human reproductive fitness. A comm... more Culture is considered an evolutionary adaptation that enhances human reproductive fitness. A common explanation is that social learning, the learning mechanism underlying cultural transmission, enhances fitness by avoiding the extra costs of individual learning. This explanation was disproved by a mathematical model of individual and social learning, showing that social learners can invade a population but do not enhance its fitness. We extend this model to include a more complex environment, limited cognitive resources, and hybrid learners that combine social and individual learning. In this extended model, we show that social learning evolves and enhances population fitness via hybrid learners capable of specializing their individual learning.
Although it is commonly assumed that rule-based models generalize more effectively than do connec... more Although it is commonly assumed that rule-based models generalize more effectively than do connectionist models, the comparison is often confounded by pitting handwritten rules against learned connections. Three case studies from cognitive development show that, under the constraint that both types of models learn their representations from equivalent examples, generalization is consistently superior in connectionist models.
Autonomous learning is the ability to learn effectively without much external assistance, which i... more Autonomous learning is the ability to learn effectively without much external assistance, which is a desirable characteristic in both engineering and computational-modeling. We extend a constructive neural-learning algorithm, sibling-descendant cascade-correlation, to monitor lack of progress in learning in order to autonomously abandon unproductive learning. The extended algorithm simulates results of recent experiments with infants who abandon learning on difficult tasks. It also avoids network overtraining effects in a more realistic manner than conventional use of validation test sets. Some contributions and limitations of constructive neural networks for achieving autonomy in learning are briefly assessed.
We propose a constructive neural-network model comprised of deterministic units which estimates a... more We propose a constructive neural-network model comprised of deterministic units which estimates and represents probability distributions from observable events -- a phenomenon related to the concept of probability matching. We use a form of operant learning, where the underlying probabilities are learned from positive and negative reinforcements of the inputs. Our model is psychologically plausible because, similar to humans, it learns to represent probabilities without receiving any representation of them from the external world, but rather by experiencing individual events. Also, we discuss how the estimated probabilities can be used in a setting with deterministic units to produce matching behaviour in choice. Our work is a step towards understanding the neural mechanisms underlying probability matching behavior by specifying processes at the algorithmic level.
Although the reasons for the success of computer simulations of psychology are often difficult to... more Although the reasons for the success of computer simulations of psychology are often difficult to identify, it is possible to make some progress through systematic experimentation. Reasons for the success of cascadecorrelation models of cognitive development are identified in two case studies. Cascade-correlation is a generative neural network model that constructs its own topology as it recruits hidden units. Capturing correct stage sequences in the integration of velocity, time, and distance cues requires a system that grows in computational power while it learns. Static networks are either too weak or too powerful to capture the full range of stages. Simulating the variation and stages in acquisition of the semantics of English personal pronouns requires sensitivity to differing amounts of addressed and non-addressed speech. Just as with children, networks benefit from the opportunity to hear personal pronouns used in exchanges between o ther speakers. Other simulations suggest t...
Learning by imitation is a powerful process for acquiring new knowledge, but there has been littl... more Learning by imitation is a powerful process for acquiring new knowledge, but there has been little research exploring imitation’s potential in the problem solving domain. Classical problem solving techniques tend to center around reinforcement learning, which requires significant trial-and-error learning to reach successful goals and problem solutions. Heuristics, hints, and reasoning by analogy have been favored as improvements over reinforcement learning, whereas imitation learning has been regarded as rote memorizing. However, research on imitation learning in animals and infants suggests that what is being learned is the overall arrangement of actions (sequencing and planning). Applied to problem solving, this suggests that imitation learning might enable a problem solver to infer a complex hierarchical problem representation from observation alone. We compared three types of learning in problem solving tasks: imitation learning (a group that viewed successful problem solving de...
Learning the Structure of Abstract Groups Dirk Schlimm ([email protected]) Department of Phi... more Learning the Structure of Abstract Groups Dirk Schlimm ([email protected]) Department of Philosophy and School of Computer Science, McGill University 855 Sherbrooke St. W., Montreal, QC H3A 2T7, Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University 1205 Penfield Avenue, Montreal, QC H3A 1B1, Canada Abstract It has recently been shown that neural networks can learn particular mathematical groups, for example, the Klein 4- group (Jamrozik & Shultz, 2007). However, there are groups with any number of elements, all of which are said to instantiate the abstract group structure. Learning to differentiate groups from other structures that are not groups is a very difficult task. Contrary to some views, we show that neural networks can learn to recognize finite groups consisting of up to 4 elements. We present this problem as a case study that exhibits the advantages of knowledge-based learning over knowledge-free l...
The Cascade-Correlation connectionist architecture was used to model human cognitive development ... more The Cascade-Correlation connectionist architecture was used to model human cognitive development on balance scale problems. The simulations were characterized by gradual expansion of the training patterns, training bias in favor of equal distance problems, and test problems balanced for torque difference. Both orderly rule stages and torque difference effects were obtained. Analyses of the development of network structure revealed progressive sensitivity to distance information. It was noted that information salience effects, such as that for torque difference, are particularly difficult to capture in symbolic level models.
A knowledge-based constructive learning algorithm, KBCC, simplifies and accelerates the learning ... more A knowledge-based constructive learning algorithm, KBCC, simplifies and accelerates the learning of parity and chessboard problems. Previously learned knowledge of simpler versions of these problems is recruited in the service of learning more complex versions. A learned solution can be viewed as a composition in which the components are not altered, showing that concatenative compositionality can be achieved in neural terms. The idea of compositionality is that mental representations are built out of parts and possess a meaning that is derived from the meanings of the parts and how the parts are combined. A symbolic expression can exhibit what is called concatenative compositionality, which means that the expression incorporates its constituents without changing them. This kind of compositionality makes it possible for symbolic propositions to express the hierarchical, tree-like structure of sentences in a natural language. It has been argued that compositionality exists in a wide ...
Neural-network Modelling of Bayesian Learning and Inference Milad Kharratzadeh (milad.kharratzade... more Neural-network Modelling of Bayesian Learning and Inference Milad Kharratzadeh ([email protected]) Department of Electrical and Computer Engineering, McGill University, 3480 University Street Montreal, QC H3A 2A7 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract We propose a modular neural-network structure for imple- menting the Bayesian framework for learning and inference. Our design has three main components, two for computing the priors and likelihoods based on observations and one for apply- ing Bayes’ rule. Through comprehensive simulations we show that our proposed model succeeds in implementing Bayesian learning and inference. We also provide a novel explanation of base-rate neglect, the most well-documented deviation from Bayes’ rule, by modelling it as a weight decay mechanism which increases entropy. Keywords: Neural-networ...
Contrary to a recent claim that neural network models are unable to account for data on infant ha... more Contrary to a recent claim that neural network models are unable to account for data on infant habituation to artificial language sentences, the present simulations show successful coverage with cascade-correlation networks using analog encoding. The results demonstrate that a symbolic rule-based account is not required by the infant data. One of the fundamental issues of cognitive science continues to revolve around which type of theoretical model better accounts for human cognition -- a symbolic rulebased account or a sub-symbolic neural network account. A recent study of infant habituation to expressions in an artificial language claims to have struck a damaging blow to the neural network approach (Marcus, Vijayan, Rao, & Vishton, 1999). The results of their study show that 7month-old infants attend longer to sentences with unfamiliar structures than to sentences with familiar structures. Because of certain features of their experimental design and their own unsuccessful neural n...
The non-linear complexities of neural networks make network solutions difficult to understand. Sa... more The non-linear complexities of neural networks make network solutions difficult to understand. Sanger's contribution analysis is here extended to the analysis of networks automatically generated by the cascade-correlation learning algorithm. Because such networks have cross connections that supersede hidden layers, standard analyses of hidden unit activation patterns are insufficient. A contribution is defined as the product of an output weight and the associated activation on the sending unit, whether that sending unit is an input or a hidden unit, multiplied by the sign of the output target for the current input pattern. Intercorrelations among contributions, as gleaned from the matrix of contributions x input patterns, can be subjected to principal components analysis (PCA) to extract the main features of variation in the contributions. Such an analysis is applied to three problems, continuous XOR, arithmetic comparison, and distinguishing between two interlocking spirals. In...
Development of Prototype Abstraction and Exemplar Memorization Irina Baetu ([email protected]... more Development of Prototype Abstraction and Exemplar Memorization Irina Baetu ([email protected]) Department of Psychology, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada likely to rely on prototypes at the beginning of a categorization task, and as training progresses they rely more on memorized exemplars (Horst, Oakes, & Madole, 2005; Minda & Smith, 2001; Smith & Minda, 1998). These studies are consistent with a shift from early prototype use to later exemplar memorization. In addition to the amount of experience with a categorization task, category structure also influences which type of information is most used. Better-structured categories can be represented as separate clusters in psychological space, whereas poorly structured categories overlap with each other (Figure 1). Smith and Mi...
Strategies, Heuristics and Biases in Complex Problem Solving Frederic Dandurand ([email protected]... more Strategies, Heuristics and Biases in Complex Problem Solving Frederic Dandurand ([email protected]) Department of Psychology, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada Kristine H. Onishi ([email protected]) Department of Psychology, McGill University 1205 Dr. Penfield Avenue, Montreal, Quebec, H3A 1B1, Canada major blocks or sets during preliminary analyses that may limit participants’ ability to explore promising areas of the problem space: a symmetry bias, a simplicity bias, and a tendency to ignore some important aspect of the task. We used a complex puzzle involving the weighing of balls. Participants had to find, with three uses of a scale, the one ball that was either heavier or lighter than the rest of a set of 12 balls. Any combination of balls was a...
Learning the Structure of a Mathematical Group Anna Jamrozik ([email protected]) Depar... more Learning the Structure of a Mathematical Group Anna Jamrozik ([email protected]) Department of Psychology, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract The Klein 4-group A mathematical group is a set of operations that satisfies the properties of identity, inverse, associativity, and closure. Understanding of group structure can be assessed by changing elements and operations over different versions of the same underlying group. Participants learned this structure more quickly over four different, successive versions of a subset of the Klein 4-group, suggesting some understanding of the group structure (Halford, Bain, Mayberry, & Andrews, 1998). Because an artificial neural network learning the task failed to improve, it was argued that such models are incapable of learning a...
Many of the problems of children's cognitive development have been difficult and longstanding... more Many of the problems of children's cognitive development have been difficult and longstanding. Among these are issues of how children represent knowledge, whether stages exist, and how children achieve transitions between stages. Despite a century of scientific evidence on child development, comprehensive theoretical understanding of these issues has remained elusive. Part of the reason is that the problems of psychological development are too complex for traditional verbal theories of development. Nonetheless, it has become clear in recent years that considerable leverage on these problems can be gained by applying computational modeling, particularly using artificial neural networks. This is because computational modeling is a good way to capture complex processes, neural networks capture developmental phenomena in a natural way, and successful models can be examined in detail to discover insights into the phenomena that they simulate (Shultz, 2001).
Recent experiments reveal that 6- to 12-month-old infants can learn probabilities and reason with... more Recent experiments reveal that 6- to 12-month-old infants can learn probabilities and reason with them. In this work, we present a novel computational system called Neural Probability Learner and Sampler (NPLS) that learns and reasons with probabilities, providing a computationally sufficient mechanism to explain infant probabilistic learning and inference. In 24 computer simulations, NPLS shows how probability distributions can emerge naturally from neural-network learning of event sequences, providing a novel explanation of infant probabilistic learning and reasoning. Three mathematical proofs show how and why NPLS simulates the infant results so accurately. The results are situated in relation to seven other active research lines. This work provides an effective way to integrate Bayesian and neural-network approaches to cognition. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
Assessing Generalization in Connectionist and Rule-based Models Under the Learning Constraint Tho... more Assessing Generalization in Connectionist and Rule-based Models Under the Learning Constraint Thomas R. Shultz ([email protected]) Department of Psychology; McGill University Montreal, QC H3C 1B1 Canada Abstract Although it is commonly assumed that rule-based models generalize more effectively than do connectionist models, the comparison is often confounded by pitting hand-written rules against learned connections. Three case studies from cognitive development show that, under the constraint that both types of models learn their representations from equivalent examples, generalization is consistently superior in connectionist models. Generalization Problems A significant part of the ongoing debate between rule- based and connectionist modeling in psychology has focused on the ability to generalize. A common claim from supporters of the classical, symbolic approach is that rule-based models are superior because they generalize more effectively than do connectionist models (Ling ...
A Fresh Look at Vocabulary Spurts Frederic Dandurand ([email protected]) Department of... more A Fresh Look at Vocabulary Spurts Frederic Dandurand ([email protected]) Department of Psychology, Universite de Montreal, 90 ave. Vincent-d'Indy Montreal, QC H2V 2S9 Canada Thomas R. Shultz ([email protected]) Department of Psychology and School of Computer Science, McGill University, 1205 Penfield Avenue Montreal, QC H3A 1B1 Canada Abstract There is currently rather little agreement about the existence of, and explanation for, a vocabulary spurt in children during the second year. Here we apply a Functional Data Analysis- based technique called Automatic Maxima Detection to the problem of finding vocabulary spurts in a sample of 20 children. Even with considerable smoothing of the data, children were found to exhibit multiple vocabulary spurts of varying intensity and location. These results should provide a clearer target for researchers interested in detecting and explaining these deviations from linear growth. Keywords: vocabulary spurts; functional data an...
Culture is considered an evolutionary adaptation that enhances human reproductive fitness. A comm... more Culture is considered an evolutionary adaptation that enhances human reproductive fitness. A common explanation is that social learning, the learning mechanism underlying cultural transmission, enhances fitness by avoiding the extra costs of individual learning. This explanation was disproved by a mathematical model of individual and social learning, showing that social learners can invade a population but do not enhance its fitness. We extend this model to include a more complex environment, limited cognitive resources, and hybrid learners that combine social and individual learning. In this extended model, we show that social learning evolves and enhances population fitness via hybrid learners capable of specializing their individual learning.
Uploads
Papers by Thomas Shultz