Results 1  10
of
18
Statistical Control of RBFlike Networks for Classification
 In 7th International Conference on Artificial Neural Networks
, 1997
"... . Incremental Net Pro (IncNet Pro) with local learning feature and statistically controlled growing and pruning of the network is introduced. The architecture of the net is based on RBF networks. Extended Kalman Filter algorithm and its new fast version is proposed and used as learning algorithm. In ..."
Abstract

Cited by 30 (14 self)
 Add to MetaCart
. Incremental Net Pro (IncNet Pro) with local learning feature and statistically controlled growing and pruning of the network is introduced. The architecture of the net is based on RBF networks. Extended Kalman Filter algorithm and its new fast version is proposed and used as learning algorithm. IncNet Pro is similar to the Resource Allocation Network described by Platt in the main idea of the expanding the network. The statistical novel criterion is used to determine the growing point. The Biradial functions are used instead of radial basis functions to obtain more flexible network. 1 Introduction The Radial Basis Function (RBF) networks [13,12] were designed as a solution to an approximation problem in multidimensional spaces. The typical form of the RBF network can be written as f(x; w;p) = M X i=1 w i G i (jjxjj i ; p i ) (1) where M is the number of the neurons in hidden layer, G i (jjxjj i ; p i ) is the i th Radial Basis Function, p i are adjustable parameters such as...
Hierarchical BayesianKalman Models For Regularisation And ARD In Sequential Learning
 DEPARTMENT OF ENGINEERING, CAMBRIDGE UNIVERSITY
, 1998
"... In this paper, we show that a hierarchical Bayesian modelling approach to sequential learning leads to many interesting attributes such as regularisation and automatic relevance determination. We identify three inference levels within this hierarchy, namely model selection, parameter estimation and ..."
Abstract

Cited by 17 (4 self)
 Add to MetaCart
(Show Context)
In this paper, we show that a hierarchical Bayesian modelling approach to sequential learning leads to many interesting attributes such as regularisation and automatic relevance determination. We identify three inference levels within this hierarchy, namely model selection, parameter estimation and noise estimation. In environments where data arrives sequentially, techniques such as crossvalidation to achieve regularisation or model selection are not possible. The Bayesian approach, with extended Kalman filtering at the parameter estimation level, allows for regularisation within a minimum variance framework. A multilayer perceptron is used to generate the extended Kalman filter nonlinear measurements mapping. We describe several algorithms at the noise estimation level, which allow us to implement adaptive regularisation and automatic relevance determination of model inputs and basis functions. An important contribution of this paper is to show the theoretical links between adaptive...
Bayesian Methods for Neural Networks
, 1999
"... Summary The application of the Bayesian learning paradigm to neural networks results in a flexible and powerful nonlinear modelling framework that can be used for regression, density estimation, prediction and classification. Within this framework, all sources of uncertainty are expressed and meas ..."
Abstract

Cited by 12 (0 self)
 Add to MetaCart
(Show Context)
Summary The application of the Bayesian learning paradigm to neural networks results in a flexible and powerful nonlinear modelling framework that can be used for regression, density estimation, prediction and classification. Within this framework, all sources of uncertainty are expressed and measured by probabilities. This formulation allows for a probabilistic treatment of our a priori knowledge, domain specific knowledge, model selection schemes, parameter estimation methods and noise estimation techniques. Many researchers have contributed towards the development of the Bayesian learning approach for neural networks. This thesis advances this research by proposing several novel extensions in the areas of sequential learning, model selection, optimisation and convergence assessment. The first contribution is a regularisation strategy for sequential learning based on extended Kalman filtering and noise estimation via evidence maximisation. Using the expectation maximisation (EM) algorithm, a similar algorithm is derived for batch learning. Much of the thesis is, however, devoted to Monte Carlo simulation methods. A robust Bayesian method is proposed to estimate,
Using affine correspondence to estimate 3d facial pose
 In Proc. Intl. Conf. Image Processing
, 2001
"... ..."
(Show Context)
Approximation and Classification in Medicine with IncNet Neural Networks
 IN MACHINE LEARNING AND APPLICATIONS. WORKSHOP ON MACHINE LEARNING IN MEDICAL APPLICATIONS
, 1999
"... Structure of incremental neural network (IncNet) is controlled by growing and pruning to match the complexity of training data. Extended Kalman Filter algorithm and its fast version is used as learning algorithm. Bicentral transfer functions, more flexible than other functions commonly used in arti ..."
Abstract

Cited by 7 (1 self)
 Add to MetaCart
Structure of incremental neural network (IncNet) is controlled by growing and pruning to match the complexity of training data. Extended Kalman Filter algorithm and its fast version is used as learning algorithm. Bicentral transfer functions, more flexible than other functions commonly used in artificial neural networks, are used. The latest improvement added is the ability to rotate the contours of constant values of transfer functions in multidimensional spaces with only N  1 adaptive parameters. Results on approximation benchmarks and on the real world psychometric classification problem clearly shows superior generalization performance of presented network comparing with other classification models.
RF free ultrasonic positioning
 In Seventh International Symposium on Wearable Computers
, 2003
"... All wearable centric location sensing technologies must address the issue of clock synchronisation between signal transmitting systems and signal receiving systems. GPS receivers, for example, compensate for synchronisation errors by incorporating a model of the receiver clock offset in the navigati ..."
Abstract

Cited by 7 (4 self)
 Add to MetaCart
All wearable centric location sensing technologies must address the issue of clock synchronisation between signal transmitting systems and signal receiving systems. GPS receivers, for example, compensate for synchronisation errors by incorporating a model of the receiver clock offset in the navigation solution. Drift between satellite clocks is also monitored to keep signal data in synch with GPS time. Most ultrasonic positioning systems solve the synchronisation problem by using a second medium for communication between transmitter and receiver devices. The transmitters in these systems emit RF signals (pings) to indicate the transmission of subsequent ultrasound signals (chirps). By subtracting the arrival time of the ping from that of the chirps, the receiver is able to compute the distance to each transmitter. In this paper, we describe an ultrasonic positioning system that does not use RF signals to achieve synchronisation. Instead, it exploits a periodic chirp transmission pattern to model the receiver’s position using chirp reception times exclusively. Not only does the system improve on the accuracy of previous technologies but it also eliminates bulky RF circuitry – a definite advantage for wearable applications where component size and weight are critical for usability. 1
Statistical Control Of Growing And Pruning In RBFLike Neural Networks
 IN THIRD CONFERENCE ON NEURAL NETWORKS AND THEIR APPLICATIONS
, 1997
"... Incremental Net Pro (IncNet Pro) with local learning feature and statistically controlled growing and pruning of the network is introduced. The architecture of the net is basedon RBF networks. Extended Kalman Filter algorithm and its new fast version is proposed and used as learning algorithm. Inc ..."
Abstract

Cited by 5 (4 self)
 Add to MetaCart
(Show Context)
Incremental Net Pro (IncNet Pro) with local learning feature and statistically controlled growing and pruning of the network is introduced. The architecture of the net is basedon RBF networks. Extended Kalman Filter algorithm and its new fast version is proposed and used as learning algorithm. IncNet Pro is similar to the Resource Allocation Network described by Platt in the main idea of the growing. The statistical novel criterion is used to determine the growing point. IncNet Pro use pruning method similar to Optimal Brain Surgeon by Hassibi, but based on Extended Kalman Filter algorithm. The Biradial functions are used instead of radial basis functions to obtain more flexible network.
Face Tracking And Pose Estimation Using Affine Motion Parameters
"... We describe a method for tracking a person's face through an image sequence and estimating the 3D facial pose within each frame. The technique is based on an affine approximation to the motion of projected facial features such as eyes, mouth and nose. Tracking stability is maintained by enforc ..."
Abstract

Cited by 5 (0 self)
 Add to MetaCart
(Show Context)
We describe a method for tracking a person's face through an image sequence and estimating the 3D facial pose within each frame. The technique is based on an affine approximation to the motion of projected facial features such as eyes, mouth and nose. Tracking stability is maintained by enforcing the affine relationship amongst the motion of the features using linear regression and application of a Kalman filter to the estimated affine parameters. Facial pose is estimated using an ellipsecircle correspondence technique based on the affine transformation between the features in the current view and those in a frontoparallel view. The method has the advantage of being simple to implement and not relying on assumed facial characteristics. Experiments on both synthetic and real sequences illustrate the effectiveness of the approach.
Estimating the Structure of Textured Surfaces Using Local Affine Flow
 in Proc. of BMVC00, 2000
, 2000
"... This paper describes a novel approach for recovering the structure and motion of a rigid textured surface from an image sequence. Camera focal length is also recovered, yielding metric estimates of the structure without the need for precalibration. The key innovation is the use of local affine flow ..."
Abstract

Cited by 2 (0 self)
 Add to MetaCart
(Show Context)
This paper describes a novel approach for recovering the structure and motion of a rigid textured surface from an image sequence. Camera focal length is also recovered, yielding metric estimates of the structure without the need for precalibration. The key innovation is the use of local affine flow parameters as the measurements within an extended Kalman filter (EKF) estimation framework, in contrast to feature correspondences or optical flow used in previous approaches. This enables surface normals to be recovered in addition to depth, unlike a feature correspondence scheme, but without the computational limitation of an optical flow approach. The method is based on equating the affine parameters to a local linearisation of the 2D motion field and using the EKF to provide recursive estimates of the 3D structure and motion. Experiments on both synthetic and real sequences demonstrate that the approach has considerable potential. 1
Uncalibrated Narrow Baseline Augmented Reality
"... We describe initial work on a system for augmenting video sequences with 3D graphics so that they appear to be present within the scene. Our aim is to do this in realtime for sequences captured by uncalibrated `live' cameras, such as a handheld or wearable. The paper focuses on obtaining 3D ..."
Abstract

Cited by 2 (1 self)
 Add to MetaCart
(Show Context)
We describe initial work on a system for augmenting video sequences with 3D graphics so that they appear to be present within the scene. Our aim is to do this in realtime for sequences captured by uncalibrated `live' cameras, such as a handheld or wearable. The paper focuses on obtaining 3D camera motion and depth estimates for these types of sequences using sparse feature tracking and the recursive algorithm developed by Azarbayejani and Pentland [1]. We report experiments which demonstrate that the approach performs well and discuss implementation issues relating to its use in a `live' realtime system.