Michael Denham - Profile on Academia.edu (original) (raw)
Papers by Michael Denham
A synaptic learning rule based on the temporal coincidence of pre- and postsynaptic activity
IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222)
ABSTRACT
Reviews in the Neurosciences, 1999
We discuss the role of the hippocampus in information processing in the brain and hypothesise tha... more We discuss the role of the hippocampus in information processing in the brain and hypothesise that the hippocampus monitors the stability of sensory cues it receives from the external world, using the current context to predict the next sensory event in the episodic sequence by learning from experience, and memorising these sequences of sensory events. Two computational models are presented here. The predictive theory and model are closely related to experimental evidence and use dynamic synapses with an asymmetric learning rule to develop predictive neural activity of a leaky integrate-and-fire model of a pyramidal CA3 cell. The oscillatory model of the hippocampus for memorising sequences of sensory events is developed as a chain of interacting neural oscillators forced by oscillatory inputs from the entorhinal cortex and from the medial septum.
The superiority of the human brain in information retrieval (IR) tasks seems to come firstly from... more The superiority of the human brain in information retrieval (IR) tasks seems to come firstly from its ability to read and understand the concepts, ideas or meanings central to documents, in order to reason out the usefulness of documents to information needs, and secondly from its ability to learn from experience and be adaptive to the environment. In this work we attempt to incorporate these properties into the development of an IR model to improve document retrieval. We investigate the applicability of concept lattices, which are based on the theory of Formal Concept Analysis (FCA), to the representation of documents. This allows the use of more elegant representation units, as opposed to keywords, in order to better capture concepts/ideas expressed in natural language text. We also investigate the use of a reinforcement leaming strategy to learn and improve document representations, based on the information present in query statements and user relevance feedback. Features or conc...
Understanding the neocortical neural architecture and circuitry in the brain that subserves our p... more Understanding the neocortical neural architecture and circuitry in the brain that subserves our perceptual and cognitive abilities will be an important component of a “Grand Challenge” which aims at an understanding of the architecture of mind and brain. We have recently embarked on a new five-year collaborative research programme, the primary aim of which is to build a computational model of minimal complexity that captures the fundamental information processing properties of the laminar microcircuitry of the primary visual area of neocortex. Specifically the properties we aim to capture are those of self-organisation, adaptation, and plasticity, which would enable the model to: (i). develop feature selective neuronal properties and cortical preference maps in response to a combination of intrinsic, spontaneously-generated activity and complex naturalistic external stimuli; and (ii) display experience-dependent and adaptation-induced plasticity, which optimally modifies the feature...
The model of novelty detection based on frequency coding of information
A model of the perception of concurrent vowels at short durations
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
The perception of double vowel stimuli has previously been modelled on the basis of determining t... more The perception of double vowel stimuli has previously been modelled on the basis of determining the pitch of the component sounds, using this information to segregate frequency channels which comprise each of the sounds, and then using a "vowel" template to identify the stimulus. Such models fail to adequately account for the human ability to distinguish vowels with the same fundamental frequency, or the effects of vowel inharmonicity, and have difficulty in explaining the results of experiments using short duration stimuli. The model we describe is based on the connectivity of the thalamocortical system and shows how a competitive recognition process and topdown suppression of the dominant vowel can lead to the identification of both vowels even in the absence of pitch grouping cues. This model provides a novel explanation for the perception of double vowel stimuli and is consistent with results showing the dependence of the phenomenon on stimulus duration.
This paper describes a neural implementation of the resistive grid technique for route finding. T... more This paper describes a neural implementation of the resistive grid technique for route finding. The resistive grid, or Laplacian planning technique, is not plagued by local minima problems and guaranties and existing route to be found. The neural network comprises 2 layers. In the upper layer, lateral connections between neurons communicate information on the potentials of neighbouring nodes in the grid. The lower layer represents a spatial memory in which information on the positions of obstacles an the target is stored. Each neuron in the upper layer receives a single input from a node in the lower layer corresponding to the same spatial location. This input is used to constrain the potentials in selected nodes in the resistive grid. The interplay between resistive grid and spatial memory results in a very flexible architecture easily adaptable to new environments. Its properties are demonstrated with a 2-dimensional path-planning problem for a mobile robot. The Dirichlet and Neumann boundary conditions are compared in terms of routes found and computational costs. Limitations and possible developments of the resistive grid technique are discussed.
Survey of discrete-event control
Major aspects of current research in the field of discrete event control system design are identi... more Major aspects of current research in the field of discrete event control system design are identified, and the requirements for future research, in particular related to CAD methods and tools, are discussed. Existing modelling methods for discrete and hybrid systems are summarised, then the establishment of a control theory and the derivation of controllers is discussed. The properties of hybrid and discrete event models include the ability to model control software directly.
Neurodynamics Research Group
A thalamocortical model of auditory streaming
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
The auditory system segregates incoming acoustic signals into perceptual representations of sound... more The auditory system segregates incoming acoustic signals into perceptual representations of sound sources within the environment, but how and where this is done is not yet clear. Psychophysical experiments on auditory streaming provide many clues about processing within the auditory system and offer a good basis for the development of models of auditory processing within a simplified paradigm. We propose a physiological basis for the process of primitive streaming, exploring our ideas by means of a computational model of the system. The model provides a novel explanation of the formation of auditory streams, and a basis for the integration of other grouping cues and attention.
An oscillatory model of sparse distributed memory and novelty detection
Biosystems, 2000
Oscillatory neural networks and their application to sensory-motor coordination and control in adaptive robots
The authors research programme has concentrated on two areas: (i) investigation of the dynamic be... more The authors research programme has concentrated on two areas: (i) investigation of the dynamic behaviour of networks composed of arrays of coupled nonlinear differential equations, each equation modelling the leaky integrator shunting dynamics of membrane potential; (ii) the development of an initial outline scheme for a sensory-motor coordination and control system for an intelligent robot based on oscillatory neural networks and synchronous association mechanisms. The authors discuss the background to the research and then present their preliminary results of simulation experiments (synchronised oscillations and chaotic behaviour) and of coordination and control system. >
Sensory-motor control of intelligent robots using oscillatory neural networks
1991 Proceedings of the 30th Ieee Conference on Decision and Control, Dec 11, 1991
The authors describe some preliminary results on the application of oscillatory neural networks t... more The authors describe some preliminary results on the application of oscillatory neural networks to the coordination and control of nonrhythmic, sensor-induced, free movement in intelligent adaptive industrial robots. In such devices, as appears to be the case in living creatures, sensory stimuli arrive at specific sensors and generate oscillatory space-time neural activity which results in recognition, association, and generation of
A model of predictive learning in the rat hippocampal principal cells during spatial activity
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
We propose a model of predictive learning in the rat hippocampus which is based on the synaptic p... more We propose a model of predictive learning in the rat hippocampus which is based on the synaptic potentiation of hippocampal pyramidal cells during temporally asymmetric pairing of EPSPs and backpropagating action potentials. The proposal is that potentiation of the synapses that mossy fibre projections of the dentate gyrus make with pyramidal cells in the CA3 region of the hippocampus causes a gradual shift backward in time of the postsynaptic activity of the CA3 cells, the firing of which initially corresponded to the rat being in a particular location. Thus these "place cells" now fire before the rat reaches this location, causing an apparent backward shift in space of the place field of the cell. The model offers both an explanation of this phenomenon which has been observed experimentally, and a possible basis for further understanding of the role of the hippocampus.
Robot control using temporal sequence learning
A Petri net approach to discrete event control
Describes a theory and procedure for the analysis and synthesis of supervisory controllers for di... more Describes a theory and procedure for the analysis and synthesis of supervisory controllers for discrete-event systems such as computing, communication and manufacturing systems. The approach described is based on the use of a Petri net model of the discrete-event system. The approach involves the control of the logical behaviour of the system, rather than its control to meet quantitative performance measures. >
We propose an automated system for out-of-sample predictions of a set of European stock indices. ... more We propose an automated system for out-of-sample predictions of a set of European stock indices. The system performs on piecewise linear representations of the time series. An automated segmentation algorithm converges to an optimum segmented time series representation, which achieves considerable data compression and allows variable sampling rate of the time series depending on different segments having different length. The minimum embedding segment dimension (MESD) algorithm, we propose, seeks for deterministic behavior of the processing data set. MESD returns the embedding dimension of the underlying dynamics of the series, measured in number of segments. Embedding dimension calculations have never been applied on segmented representations. We summarize the advantages of the method in the following: (i) it can detect high dimensional nonlinear deterministic behavior as being projected on a lower dimension segment space; (ii) it is computationally efficient; (iii) it converges to an optimum solution without a-priori parameterization. We use the minimum embedding segment dimension (MESD) as an indicator of the length of patterns that can be retrieved from the time series own past. Our pattern matching technique enables the matching of such historical patterns on others of similar shape which occur in different time scales. To define an appropriate similarity measure, we introduce the notation of Multiple Feature Sets (MFS) which employ Dynamic Time Warping (DTW), and first derivative and temporal features. An additional advantage of the system we propose is that the segmented representation scheme and the prediction model are both data driven and that the predictions are made using information only from the time-series own past without any a priori knowledge being injected into the model. We demonstrate that this approach may offer a useful decision support tool for stock market trading.
Control of discrete event processes
1986 25th Ieee Conference on Decision and Control, 1986
This paper considers the problem of devising an adequate theory in which the behaviour of flexibl... more This paper considers the problem of devising an adequate theory in which the behaviour of flexible manufacturing systems, at the level of coordination and control of a set of interacting, communicating concurrent processes, can be modelled and analysed. Existing approaches to a solution are reviewed and the outline of a proposed new approach based on some recent concepts such as object oriented systems is described.
In this paper, we propose a model of predictive learning in the septal-hippocampal region of the ... more In this paper, we propose a model of predictive learning in the septal-hippocampal region of the brain. We suggest that this region plays a primary role in monitoring the stability of the sensory cues that an animal receives from the external world, in relation to its own internally generated knowledge of its location, direction of gaze, and direction of movement. Our predictive model of information processing is based on the synaptic potentiation of hippocampal pyramidal cells during temporally asymmetric pairing of EPSPs and back-propagating action potentials. The model demonstrates that such potentiation of the synapses that the mossy fibre projections from the dentate gyrus make with pyramidal cells in the CA3 region of the hippocampus can cause a gradual shift backward in time of the postsynaptic activity of the CA3 cells. thus making their activity predictive of the upcoming sensory state of the animal. This is predictive activity has been observed experimentally in the firing of "place" cells in rat hippocampus. which corresponds initially to the rat being in a particular location. After repetitive experience, these cells begin to fire before the rat reaches this location, causing an apparent backward shift in space of the "place" field of the cell. The model offers both an explanation of this phenomenon, and more generally, a possible basis for further understanding of the role of the septal-hippocampal region in learning and memory.
Concept Based Adaptive IR Model Using FCA-BAM Combination for Concept Representation and Encoding
... Centre for Neural and Adaptive Systems, University of Plymouth, UK rohan/mike@soc.plym.ac. uk... more ... Centre for Neural and Adaptive Systems, University of Plymouth, UK rohan/mike@soc.plym.ac. uk ... Amongst are the one proposed by Kosko[15], the originator of BAMs and that proposed byRadim B lohlávek [2]. A real threshold i x ( j y) is assigned to the ith neuron of the first ...
A synaptic learning rule based on the temporal coincidence of pre- and postsynaptic activity
IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222)
ABSTRACT
Reviews in the Neurosciences, 1999
We discuss the role of the hippocampus in information processing in the brain and hypothesise tha... more We discuss the role of the hippocampus in information processing in the brain and hypothesise that the hippocampus monitors the stability of sensory cues it receives from the external world, using the current context to predict the next sensory event in the episodic sequence by learning from experience, and memorising these sequences of sensory events. Two computational models are presented here. The predictive theory and model are closely related to experimental evidence and use dynamic synapses with an asymmetric learning rule to develop predictive neural activity of a leaky integrate-and-fire model of a pyramidal CA3 cell. The oscillatory model of the hippocampus for memorising sequences of sensory events is developed as a chain of interacting neural oscillators forced by oscillatory inputs from the entorhinal cortex and from the medial septum.
The superiority of the human brain in information retrieval (IR) tasks seems to come firstly from... more The superiority of the human brain in information retrieval (IR) tasks seems to come firstly from its ability to read and understand the concepts, ideas or meanings central to documents, in order to reason out the usefulness of documents to information needs, and secondly from its ability to learn from experience and be adaptive to the environment. In this work we attempt to incorporate these properties into the development of an IR model to improve document retrieval. We investigate the applicability of concept lattices, which are based on the theory of Formal Concept Analysis (FCA), to the representation of documents. This allows the use of more elegant representation units, as opposed to keywords, in order to better capture concepts/ideas expressed in natural language text. We also investigate the use of a reinforcement leaming strategy to learn and improve document representations, based on the information present in query statements and user relevance feedback. Features or conc...
Understanding the neocortical neural architecture and circuitry in the brain that subserves our p... more Understanding the neocortical neural architecture and circuitry in the brain that subserves our perceptual and cognitive abilities will be an important component of a “Grand Challenge” which aims at an understanding of the architecture of mind and brain. We have recently embarked on a new five-year collaborative research programme, the primary aim of which is to build a computational model of minimal complexity that captures the fundamental information processing properties of the laminar microcircuitry of the primary visual area of neocortex. Specifically the properties we aim to capture are those of self-organisation, adaptation, and plasticity, which would enable the model to: (i). develop feature selective neuronal properties and cortical preference maps in response to a combination of intrinsic, spontaneously-generated activity and complex naturalistic external stimuli; and (ii) display experience-dependent and adaptation-induced plasticity, which optimally modifies the feature...
The model of novelty detection based on frequency coding of information
A model of the perception of concurrent vowels at short durations
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
The perception of double vowel stimuli has previously been modelled on the basis of determining t... more The perception of double vowel stimuli has previously been modelled on the basis of determining the pitch of the component sounds, using this information to segregate frequency channels which comprise each of the sounds, and then using a "vowel" template to identify the stimulus. Such models fail to adequately account for the human ability to distinguish vowels with the same fundamental frequency, or the effects of vowel inharmonicity, and have difficulty in explaining the results of experiments using short duration stimuli. The model we describe is based on the connectivity of the thalamocortical system and shows how a competitive recognition process and topdown suppression of the dominant vowel can lead to the identification of both vowels even in the absence of pitch grouping cues. This model provides a novel explanation for the perception of double vowel stimuli and is consistent with results showing the dependence of the phenomenon on stimulus duration.
This paper describes a neural implementation of the resistive grid technique for route finding. T... more This paper describes a neural implementation of the resistive grid technique for route finding. The resistive grid, or Laplacian planning technique, is not plagued by local minima problems and guaranties and existing route to be found. The neural network comprises 2 layers. In the upper layer, lateral connections between neurons communicate information on the potentials of neighbouring nodes in the grid. The lower layer represents a spatial memory in which information on the positions of obstacles an the target is stored. Each neuron in the upper layer receives a single input from a node in the lower layer corresponding to the same spatial location. This input is used to constrain the potentials in selected nodes in the resistive grid. The interplay between resistive grid and spatial memory results in a very flexible architecture easily adaptable to new environments. Its properties are demonstrated with a 2-dimensional path-planning problem for a mobile robot. The Dirichlet and Neumann boundary conditions are compared in terms of routes found and computational costs. Limitations and possible developments of the resistive grid technique are discussed.
Survey of discrete-event control
Major aspects of current research in the field of discrete event control system design are identi... more Major aspects of current research in the field of discrete event control system design are identified, and the requirements for future research, in particular related to CAD methods and tools, are discussed. Existing modelling methods for discrete and hybrid systems are summarised, then the establishment of a control theory and the derivation of controllers is discussed. The properties of hybrid and discrete event models include the ability to model control software directly.
Neurodynamics Research Group
A thalamocortical model of auditory streaming
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
The auditory system segregates incoming acoustic signals into perceptual representations of sound... more The auditory system segregates incoming acoustic signals into perceptual representations of sound sources within the environment, but how and where this is done is not yet clear. Psychophysical experiments on auditory streaming provide many clues about processing within the auditory system and offer a good basis for the development of models of auditory processing within a simplified paradigm. We propose a physiological basis for the process of primitive streaming, exploring our ideas by means of a computational model of the system. The model provides a novel explanation of the formation of auditory streams, and a basis for the integration of other grouping cues and attention.
An oscillatory model of sparse distributed memory and novelty detection
Biosystems, 2000
Oscillatory neural networks and their application to sensory-motor coordination and control in adaptive robots
The authors research programme has concentrated on two areas: (i) investigation of the dynamic be... more The authors research programme has concentrated on two areas: (i) investigation of the dynamic behaviour of networks composed of arrays of coupled nonlinear differential equations, each equation modelling the leaky integrator shunting dynamics of membrane potential; (ii) the development of an initial outline scheme for a sensory-motor coordination and control system for an intelligent robot based on oscillatory neural networks and synchronous association mechanisms. The authors discuss the background to the research and then present their preliminary results of simulation experiments (synchronised oscillations and chaotic behaviour) and of coordination and control system. >
Sensory-motor control of intelligent robots using oscillatory neural networks
1991 Proceedings of the 30th Ieee Conference on Decision and Control, Dec 11, 1991
The authors describe some preliminary results on the application of oscillatory neural networks t... more The authors describe some preliminary results on the application of oscillatory neural networks to the coordination and control of nonrhythmic, sensor-induced, free movement in intelligent adaptive industrial robots. In such devices, as appears to be the case in living creatures, sensory stimuli arrive at specific sensors and generate oscillatory space-time neural activity which results in recognition, association, and generation of
A model of predictive learning in the rat hippocampal principal cells during spatial activity
1998 Ieee International Joint Conference on Neural Networks Proceedings Ieee World Congress on Computational Intelligence, May 4, 1998
We propose a model of predictive learning in the rat hippocampus which is based on the synaptic p... more We propose a model of predictive learning in the rat hippocampus which is based on the synaptic potentiation of hippocampal pyramidal cells during temporally asymmetric pairing of EPSPs and backpropagating action potentials. The proposal is that potentiation of the synapses that mossy fibre projections of the dentate gyrus make with pyramidal cells in the CA3 region of the hippocampus causes a gradual shift backward in time of the postsynaptic activity of the CA3 cells, the firing of which initially corresponded to the rat being in a particular location. Thus these "place cells" now fire before the rat reaches this location, causing an apparent backward shift in space of the place field of the cell. The model offers both an explanation of this phenomenon which has been observed experimentally, and a possible basis for further understanding of the role of the hippocampus.
Robot control using temporal sequence learning
A Petri net approach to discrete event control
Describes a theory and procedure for the analysis and synthesis of supervisory controllers for di... more Describes a theory and procedure for the analysis and synthesis of supervisory controllers for discrete-event systems such as computing, communication and manufacturing systems. The approach described is based on the use of a Petri net model of the discrete-event system. The approach involves the control of the logical behaviour of the system, rather than its control to meet quantitative performance measures. >
We propose an automated system for out-of-sample predictions of a set of European stock indices. ... more We propose an automated system for out-of-sample predictions of a set of European stock indices. The system performs on piecewise linear representations of the time series. An automated segmentation algorithm converges to an optimum segmented time series representation, which achieves considerable data compression and allows variable sampling rate of the time series depending on different segments having different length. The minimum embedding segment dimension (MESD) algorithm, we propose, seeks for deterministic behavior of the processing data set. MESD returns the embedding dimension of the underlying dynamics of the series, measured in number of segments. Embedding dimension calculations have never been applied on segmented representations. We summarize the advantages of the method in the following: (i) it can detect high dimensional nonlinear deterministic behavior as being projected on a lower dimension segment space; (ii) it is computationally efficient; (iii) it converges to an optimum solution without a-priori parameterization. We use the minimum embedding segment dimension (MESD) as an indicator of the length of patterns that can be retrieved from the time series own past. Our pattern matching technique enables the matching of such historical patterns on others of similar shape which occur in different time scales. To define an appropriate similarity measure, we introduce the notation of Multiple Feature Sets (MFS) which employ Dynamic Time Warping (DTW), and first derivative and temporal features. An additional advantage of the system we propose is that the segmented representation scheme and the prediction model are both data driven and that the predictions are made using information only from the time-series own past without any a priori knowledge being injected into the model. We demonstrate that this approach may offer a useful decision support tool for stock market trading.
Control of discrete event processes
1986 25th Ieee Conference on Decision and Control, 1986
This paper considers the problem of devising an adequate theory in which the behaviour of flexibl... more This paper considers the problem of devising an adequate theory in which the behaviour of flexible manufacturing systems, at the level of coordination and control of a set of interacting, communicating concurrent processes, can be modelled and analysed. Existing approaches to a solution are reviewed and the outline of a proposed new approach based on some recent concepts such as object oriented systems is described.
In this paper, we propose a model of predictive learning in the septal-hippocampal region of the ... more In this paper, we propose a model of predictive learning in the septal-hippocampal region of the brain. We suggest that this region plays a primary role in monitoring the stability of the sensory cues that an animal receives from the external world, in relation to its own internally generated knowledge of its location, direction of gaze, and direction of movement. Our predictive model of information processing is based on the synaptic potentiation of hippocampal pyramidal cells during temporally asymmetric pairing of EPSPs and back-propagating action potentials. The model demonstrates that such potentiation of the synapses that the mossy fibre projections from the dentate gyrus make with pyramidal cells in the CA3 region of the hippocampus can cause a gradual shift backward in time of the postsynaptic activity of the CA3 cells. thus making their activity predictive of the upcoming sensory state of the animal. This is predictive activity has been observed experimentally in the firing of "place" cells in rat hippocampus. which corresponds initially to the rat being in a particular location. After repetitive experience, these cells begin to fire before the rat reaches this location, causing an apparent backward shift in space of the "place" field of the cell. The model offers both an explanation of this phenomenon, and more generally, a possible basis for further understanding of the role of the septal-hippocampal region in learning and memory.
Concept Based Adaptive IR Model Using FCA-BAM Combination for Concept Representation and Encoding
... Centre for Neural and Adaptive Systems, University of Plymouth, UK rohan/mike@soc.plym.ac. uk... more ... Centre for Neural and Adaptive Systems, University of Plymouth, UK rohan/mike@soc.plym.ac. uk ... Amongst are the one proposed by Kosko[15], the originator of BAMs and that proposed byRadim B lohlávek [2]. A real threshold i x ( j y) is assigned to the ith neuron of the first ...