Generalized relevance learning vector quantization

Subspace corrected relevance learning with application in neuroimaging

In machine learning, data often comes from different sources, but combining them can introduce extraneous variation that affects both generalization and interpretability. For example, we investigate the classification of neurodegenerative diseases using FDG-PET data collected from multiple neuroimaging centers. However, data collected at different centers introduces unwanted variation due to differences in scanners, scanning protocols, and processing methods. To address this issue, we propose a two-step approach to limit the influence of center-dependent variation on the classification of healthy controls and early vs. late-stage Parkinson's disease patients. First, we train a Generalized Matrix Learning Vector Quantization (GMLVQ) model on healthy control data to identify a "relevance space" that distinguishes between centers. Second, we use this space to construct a correction matrix that restricts a second GMLVQ system's training on the diagnostic problem. We evaluate the effectiveness of this approach on the real-world multi-center datasets and simulated artificial dataset. Our results demonstrate that the approach produces machine learning systems with reduced bias - being more specific due to eliminating information related to center differences during the training process - and more informative relevance profiles that can be interpreted by medical experts. This method can be adapted to similar problems outside the neuroimaging domain, as long as an appropriate "relevance space" can be identified to construct the correction matrix.

Online Deterministic Annealing for Classification and Clustering

Inherent in virtually every iterative machine learning algorithm is the problem of hyperparameter tuning, which includes three major design parameters: 1) the complexity of the model, e.g., the number of neurons in a neural network; 2) the initial conditions, which heavily affect the behavior of the algorithm; and 3) the dissimilarity measure used to quantify its performance. We introduce an online prototype-based learning algorithm that can be viewed as a progressively growing competitive-learning neural network architecture for classification and clustering. The learning rule of the proposed approach is formulated as an online gradient-free stochastic approximation algorithm that solves a sequence of appropriately defined optimization problems, simulating an annealing process. The annealing nature of the algorithm contributes to avoiding poor local minima, offers robustness with respect to the initial conditions, and provides a means to progressively increase the complexity of the learning model, through an intuitive bifurcation phenomenon. The proposed approach is interpretable, requires minimal hyperparameter tuning, and allows online control over the performance-complexity tradeoff. Finally, we show that Bregman divergences appear naturally as a family of dissimilarity measures that play a central role in both the performance and the computational complexity of the learning algorithm.

Distance metric learning based on the class center and nearest neighbor relationship

Distance metric learning has been a promising technology to improve the performance of algorithms related to distance metrics. The existing distance metric learning methods are either based on the class center or the nearest neighbor relationship. In this work, we propose a new distance metric learning method based on the class center and nearest neighbor relationship (DMLCN). Specifically, when centers of different classes overlap, DMLCN first splits each class into several clusters and uses one center to represent one cluster. Then, a distance metric is learned such that each example is close to the corresponding cluster center and the nearest neighbor relationship is kept for each receptive field. Therefore, while characterizing the local structure of data, the proposed method leads to intra-class compactness and inter-class dispersion simultaneously. Further, to better process complex data, we introduce multiple metrics into DMLCN (MMLCN) by learning a local metric for each center. Following that, a new classification decision rule is designed based on the proposed methods. Moreover, we develop an iterative algorithm to optimize the proposed methods. The convergence and complexity are analyzed theoretically. Experiments on different types of data sets including artificial data sets, benchmark data sets and noise data sets show the feasibility and effectiveness of the proposed methods.

Quantum Computing Approaches for Vector Quantization-Current Perspectives and Developments

In the field of machine learning, vector quantization is a category of low-complexity approaches that are nonetheless powerful for data representation and clustering or classification tasks. Vector quantization is based on the idea of representing a data or a class distribution using a small set of prototypes, and hence, it belongs to interpretable models in machine learning. Further, the low complexity of vector quantizers makes them interesting for the application of quantum concepts for their implementation. This is especially true for current and upcoming generations of quantum devices, which only allow the execution of simple and restricted algorithms. Motivated by different adaptation and optimization paradigms for vector quantizers, we provide an overview of respective existing quantum algorithms and routines to realize vector quantization concepts, maybe only partially, on quantum devices. Thus, the reader can infer the current state-of-the-art when considering quantum computing approaches for vector quantization.

An Approach to Growth Delimitation of Straight Line Segment Classifiers Based on a Minimum Bounding Box

Several supervised machine learning algorithms focused on binary classification for solving daily problems can be found in the literature. The straight-line segment classifier stands out for its low complexity and competitiveness, compared to well-knownconventional classifiers. This binary classifier is based on distances between points and two labeled sets of straight-line segments. Its training phase consists of finding the placement of labeled straight-line segment extremities (and consequently, their lengths) which gives the minimum mean square error. However, during the training phase, the straight-line segment lengths can grow significantly, giving a negative impact on the classification rate. Therefore, this paper proposes an approach for adjusting the placements of labeled straight-line segment extremities to build reliable classifiers in a constrained search space (tuned by a scale factor parameter) in order to restrict their lengths. Ten artificial and eight datasets from the UCI Machine Learning Repository were used to prove that our approach shows promising results, compared to other classifiers. We conclude that this classifier can be used in industry for decision-making problems, due to the straightforward interpretation and classification rates.

Identification of multi-concentration aromatic fragrances with electronic nose technology using a support vector machine

Due to the concentration effect, there is a major challenge for the electronic nose system to identify different odor samples with multiple concentrations. The development of artificial intelligence provides new ways to solve such problems. This article attempts to use support vector machine (SVM) technology to distinguish four fragrance samples with three concentrations, including roman chamomile, jasmine, lavender, and orange. The responses of these samples were collected by an 11-sensor electronic nose. After baseline correction, data smoothing, and removal of non-responsive sensors, the signals of 8 sensors were used for subsequent model analysis. Due to the concentration effect, when the primary signal intensities were used as features, the electronic nose cannot distinguish between different aroma types (accuracy less than 50%). When the normalized maximum signal intensity X_mr was used, the accuracy of the model was greatly improved. Graphic analysis and PCA showed that the normalized feature effectively eliminates the concentration effect, and appropriately reducing some sensors can enhance the ability to distinguish odors. The SVM correctly classified all 14 aromas when feeding 8 sets of data to train the radial kernel C-classification SVM. This showed that the cross-interference of the sensors was reduced, and the resolving power of the electronic nose was enhanced after the feature reduction.

The Resolved Mutual Information Function as a Structural Fingerprint of Biomolecular Sequences for Interpretable Machine Learning Classifiers

In the present article we propose the application of variants of the mutual information function as characteristic fingerprints of biomolecular sequences for classification analysis. In particular, we consider the resolved mutual information functions based on Shannon-, Rényi-, and Tsallis-entropy. In combination with interpretable machine learning classifier models based on generalized learning vector quantization, a powerful methodology for sequence classification is achieved which allows substantial knowledge extraction in addition to the high classification ability due to the model-inherent robustness. Any potential (slightly) inferior performance of the used classifier is compensated by the additional knowledge provided by interpretable models. This knowledge may assist the user in the analysis and understanding of the used data and considered task. After theoretical justification of the concepts, we demonstrate the approach for various example data sets covering different areas in biomolecular sequence analysis.

Probabilistic learning vector quantization on manifold of symmetric positive definite matrices

In this paper, we develop a new classification method for manifold-valued data in the framework of probabilistic learning vector quantization. In many classification scenarios, the data can be naturally represented by symmetric positive definite matrices, which are inherently points that live on a curved Riemannian manifold. Due to the non-Euclidean geometry of Riemannian manifolds, traditional Euclidean machine learning algorithms yield poor results on such data. In this paper, we generalize the probabilistic learning vector quantization algorithm for data points living on the manifold of symmetric positive definite matrices equipped with Riemannian natural metric (affine-invariant metric). By exploiting the induced Riemannian distance, we derive the probabilistic learning Riemannian space quantization algorithm, obtaining the learning rule through Riemannian gradient descent. Empirical investigations on synthetic data, image data , and motor imagery electroencephalogram (EEG) data demonstrate the superior performance of the proposed method.

Self-incremental learning vector quantization with human cognitive biases

Human beings have adaptively rational cognitive biases for efficiently acquiring concepts from small-sized datasets. With such inductive biases, humans can generalize concepts by learning a small number of samples. By incorporating human cognitive biases into learning vector quantization (LVQ), a prototype-based online machine learning method, we developed self-incremental LVQ (SILVQ) methods that can be easily interpreted. We first describe a method to automatically adjust the learning rate that incorporates human cognitive biases. Second, SILVQ, which self-increases the prototypes based on the method for automatically adjusting the learning rate, is described. The performance levels of the proposed methods are evaluated in experiments employing four real and two artificial datasets. Compared with the original learning vector quantization algorithms, our methods not only effectively remove the need for parameter tuning, but also achieve higher accuracy from learning small numbers of instances. In the cases of larger numbers of instances, SILVQ can still achieve an accuracy that is equal to or better than those of existing representative LVQ algorithms. Furthermore, SILVQ can learn linearly inseparable conceptual structures with the required and sufficient number of prototypes without overfitting.

Generalized Learning Riemannian Space Quantization: A Case Study on Riemannian Manifold of SPD Matrices

Learning vector quantization (LVQ) is a simple and efficient classification method, enjoying great popularity. However, in many classification scenarios, such as electroencephalogram (EEG) classification, the input features are represented by symmetric positive-definite (SPD) matrices that live in a curved manifold rather than vectors that live in the flat Euclidean space. In this article, we propose a new classification method for data points that live in the curved Riemannian manifolds in the framework of LVQ. The proposed method alters generalized LVQ (GLVQ) with the Euclidean distance to the one operating under the appropriate Riemannian metric. We instantiate the proposed method for the Riemannian manifold of SPD matrices equipped with the Riemannian natural metric. Empirical investigations on synthetic data and real-world motor imagery EEG data demonstrate that the performance of the proposed generalized learning Riemannian space quantization can significantly outperform the Euclidean GLVQ, generalized relevance LVQ (GRLVQ), and generalized matrix LVQ (GMLVQ). The proposed method also shows competitive performance to the state-of-the-art methods on the EEG classification of motor imagery tasks.

Sample-Guided Adaptive Class Prototype for Visual Domain Adaptation

Domain adaptation aims to handle the distribution mismatch of training and testing data, which achieves dramatic progress in multi-sensor systems. Previous methods align the cross-domain distributions by some statistics, such as the means and variances. Despite their appeal, such methods often fail to model the discriminative structures existing within testing samples. In this paper, we present a sample-guided adaptive class prototype method, which consists of the no distribution matching strategy. Specifically, two adaptive measures are proposed. Firstly, the modified nearest class prototype is raised, which allows more diversity within same class, while keeping most of the class wise discrimination information. Secondly, we put forward an easy-to-hard testing scheme by taking into account the different difficulties in recognizing target samples. Easy samples are classified and selected to assist the prediction of hard samples. Extensive experiments verify the effectiveness of the proposed method.

Self-organizing subspace clustering for high-dimensional and multi-view data

A surge in the availability of data from multiple sources and modalities is correlated with advances in how to obtain, compress, store, transfer, and process large amounts of complex high-dimensional data. The clustering challenge increases with the growth of data dimensionality which decreases the discriminate power of the distance metrics. Subspace clustering aims to group data drawn from a union of subspaces. In such a way, there is a large number of state-of-the-art approaches and we divide them into families regarding the method used in the clustering. We introduce a soft subspace clustering algorithm, a Self-organizing Map (SOM) with a time-varying structure, to cluster data without any prior knowledge of the number of categories or of the neural network topology, both determined during the training process. The model also assigns proper relevancies (weights) to different dimensions, capturing from the learning process the influence of each dimension on uncovering clusters. We employ a number of real-world datasets to validate the model. This algorithm presents a competitive performance in a diverse range of contexts among them data mining, gene expression, multi-view, computer vision and text clustering problems which include high-dimensional data. Extensive experiments suggest that our method very often outperforms the state-of-the-art approaches in all types of problems considered.

Adaptive basis functions for prototype-based classification of functional data

We present a framework for distance-based classification of functional data. We consider the analysis of labeled spectral data and time series by means of generalized matrix relevance learning vector quantization (GMLVQ) as an example. To take advantage of the functional nature, a functional expansion of the input data is considered. Instead of using a predefined set of basis functions for the expansion, a more flexible scheme of an adaptive functional basis is employed. GMLVQ is applied on the resulting functional parameters to solve the classification task. For comparison of the classification, a GMLVQ system is also applied to the raw input data, as well as on data expanded by a different predefined functional basis. Computer experiments show that the methods offer potential to improve classification performance significantly. Furthermore, the analysis of the adapted set of basis functions give further insights into the data structure and yields an option for a drastic reduction of dimensionality.

Atherosclerosis, myocardial infarction and primary hemostasis: Impact of platelets, von Willebrand factor and soluble glycoprotein VI

INTRODUCTION

Little is known about peril constellations in primary hemostasis contributing to an acute myocardial infarction (MI) in patients with already manifest atherosclerosis. The study aimed to establish a predicting model based on six biomarkers of primary hemostasis: platelet count, mean platelet volume, hematocrit, soluble glycoprotein VI, fibrinogen and von Willebrand factor ratio.

MATERIALS AND METHODS

The biomarkers were measured in 1.491 patients with manifest atherosclerosis of the Leipzig (LIFE) heart study. Three groups were divided: patients with coronary artery disease (900 patients) and patients with atherosclerosis and either ST-elevated MI (404 patients) or Non-ST-elevated MI (187 patients). Correlations were analyzed by non-linear analysis with Self Organizing Maps. Classification and discriminant analysis was performed using Learning Vector Quantization.

RESULTS AND CONCLUSIONS

The combination of hemostatic biomarkers is regarded as valuable tool for identifying patients with atherosclerosis at risk for MI. Nevertheless, our study contradicts this belief. The biomarkers did not allow to establish a predicting model usable in daily patient care. Good specificity and sensitivity for the detection of MI was only reached in models including acute phase parameters (specificity 0,9036, sensitivity 0,7937 in men; 0,8977 and 0,8133 in women). In detail, hematocrit and soluble glycoprotein VI were significantly different between the groups. Significant dissimilarities were also found for fibrinogen (in men) and von Willebrand factor ratio. In contrast, the most promising parameters mean platelet volume and platelet count showed no difference, which is an important contribution to the controversy concerning them as new risk and therapy targets for MI.

Application of an interpretable classification model on Early Folding Residues during protein folding

Background

Machine learning strategies are prominent tools for data analysis. Especially in life sciences, they have become increasingly important to handle the growing datasets collected by the scientific community. Meanwhile, algorithms improve in performance, but also gain complexity, and tend to neglect interpretability and comprehensiveness of the resulting models.

Results

Generalized Matrix Learning Vector Quantization (GMLVQ) is a supervised, prototype-based machine learning method and provides comprehensive visualization capabilities not present in other classifiers which allow for a fine-grained interpretation of the data. In contrast to commonly used machine learning strategies, GMLVQ is well-suited for imbalanced classification problems which are frequent in life sciences. We present a Weka plug-in implementing GMLVQ. The feasibility of GMLVQ is demonstrated on a dataset of Early Folding Residues (EFR) that have been shown to initiate and guide the protein folding process. Using 27 features, an area under the receiver operating characteristic of 76.6% was achieved which is comparable to other state-of-the-art classifiers. The obtained model is accessible at https://biosciences.hs-mittweida.de/efpred/.

Conclusions

The application on EFR prediction demonstrates how an easy interpretation of classification models can promote the comprehension of biological mechanisms. The results shed light on the special features of EFR which were reported as most influential for the classification: EFR are embedded in ordered secondary structure elements and they participate in networks of hydrophobic residues. Visualization capabilities of GMLVQ are presented as we demonstrate how to interpret the results.

Electronic supplementary material

The online version of this article (10.1186/s13040-018-0188-2) contains supplementary material, which is available to authorized users.

Characterizing the relation of functional and Early Folding Residues in protein structures using the example of aminoacyl-tRNA synthetases

Proteins are chains of amino acids which adopt a three-dimensional structure and are then able to catalyze chemical reactions or propagate signals in organisms. Without external influence, many proteins fold into their native structure, and a small number of Early Folding Residues (EFR) have previously been shown to initiate the formation of secondary structure elements and guide their respective assembly. Using the two diverse superfamilies of aminoacyl-tRNA synthetases (aaRS), it is shown that the position of EFR is preserved over the course of evolution even when the corresponding sequence conservation is small. Folding initiation sites are positioned in the center of secondary structure elements, independent of aaRS class. In class I, the predicted position of EFR resembles an ancient structural packing motif present in many seemingly unrelated proteins. Furthermore, it is shown that EFR and functionally relevant residues in aaRS are almost entirely disjoint sets of residues. The Start2Fold database is used to investigate whether this separation of EFR and functional residues can be observed for other proteins. EFR are found to constitute crucial connectors of protein regions which are distant at sequence level. Especially, these residues exhibit a high number of non-covalent residue-residue contacts such as hydrogen bonds and hydrophobic interactions. This tendency also manifests as energetically stable local regions, as substantiated by a knowledge-based potential. Despite profound differences regarding how EFR and functional residues are embedded in protein structures, a strict separation of structurally and functionally relevant residues cannot be observed for a more general collection of proteins.

Study of Face Orientation Recognition Based on Neural Network

Learning vector quantization (LVQ) network and back-propagation (BP) network are constructed easily making use of MATLAB toolbox on the basis of maintaining the recognition rate. Face images are randomly selected from images set as training data of LVQ network and BP network. LVQ algorithm and BP algorithm are used to train network. The automatic recognition of face orientation is realized when the system obtains convergence network. First, all images are processed by edge detection. Then feature vectors representing position of the eye were extracted from edge detected images. Feature vectors of training set are sent to network to adjust the parameters which ensures the convergence speed and performance of the network. Experimental results show that the constructed LVQ network and BP network can judge face orientation according to feature vectors of input images. Generally, the recognition rate of LVQ network is higher than that of BP network. The LVQ network and BP network are both feasible and effective for face orientation recognition to some extent. The advantage of this work is that the recognition system is efficient and easy to promote. This paper focuses on how to use MATLAB easily to design identification network rather than the complexity of identification system. The future research will focus on the stability and robustness of recognition network.

Spectral-spatial feature-based neural network method for acute lymphoblastic leukemia cell identification via microscopic hyperspectral imaging technology

Microscopic examination is one of the most common methods for acute lymphoblastic leukemia (ALL) diagnosis. Most traditional methods of automized blood cell identification are based on RGB color or gray images captured by light microscopes. This paper presents an identification method combining both spectral and spatial features to identify lymphoblasts from lymphocytes in hyperspectral images. Normalization and encoding method is applied for spectral feature extraction and the support vector machine-recursive feature elimination (SVM-RFE) algorithm is presented for spatial feature determination. A marker-based learning vector quantization (MLVQ) neural network is proposed to perform identification with the integrated features. Experimental results show that this algorithm yields identification accuracy, sensitivity, and specificity of 92.9%, 93.3%, and 92.5%, respectively. Hyperspectral microscopic blood imaging combined with neural network identification technique has the potential to provide a feasible tool for ALL pre-diagnosis.

Can Learning Vector Quantization be an Alternative to SVM and Deep Learning? - Recent Trends and Advanced Variants of Learning Vector Quantization for Classification Learning

Learning vector quantization (LVQ) is one of the most powerful approaches for prototype based classification of vector data, intuitively introduced by Kohonen. The prototype adaptation scheme relies on its attraction and repulsion during the learning providing an easy geometric interpretability of the learning as well as of the classification decision scheme. Although deep learning architectures and support vector classifiers frequently achieve comparable or even better results, LVQ models are smart alternatives with low complexity and computational costs making them attractive for many industrial applications like intelligent sensor systems or advanced driver assistance systems.

Nowadays, the mathematical theory developed for LVQ delivers sufficient justification of the algorithm making it an appealing alternative to other approaches like support vector machines and deep learning techniques.

This review article reports current developments and extensions of LVQ starting from the generalized LVQ (GLVQ), which is known as the most powerful cost function based realization of the original LVQ. The cost function minimized in GLVQ is an soft-approximation of the standard classification error allowing gradient descent learning techniques. The GLVQ variants considered in this contribution, cover many aspects like bordersensitive learning, application of non-Euclidean metrics like kernel distances or divergences, relevance learning as well as optimization of advanced statistical classification quality measures beyond the accuracy including sensitivity and specificity or area under the ROC-curve.

According to these topics, the paper highlights the basic motivation for these variants and extensions together with the mathematical prerequisites and treatments for integration into the standard GLVQ scheme and compares them to other machine learning approaches. For detailed description and mathematical theory behind all, the reader is referred to the respective original articles.

Thus, the intention of the paper is to provide a comprehensive overview of the stateof- the-art serving as a starting point to search for an appropriate LVQ variant in case of a given specific classification problem as well as a reference to recently developed variants and improvements of the basic GLVQ scheme.

Supervised Variational Relevance Learning, An Analytic Geometric Feature Selection with Applications to Omic Datasets

We introduce Supervised Variational Relevance Learning (Suvrel), a variational method to determine metric tensors to define distance based similarity in pattern classification, inspired in relevance learning. The variational method is applied to a cost function that penalizes large intraclass distances and favors small interclass distances. We find analytically the metric tensor that minimizes the cost function. Preprocessing the patterns by doing linear transformations using the metric tensor yields a dataset which can be more efficiently classified. We test our methods using publicly available datasets, for some standard classifiers. Among these datasets, two were tested by the MAQC-II project and, even without the use of further preprocessing, our results improve on their performance.

Self-Supervised Online Learning of Basic Object Push Affordances

Continuous learning of object affordances in a cognitive robot is a challenging problem, the solution to which arguably requires a developmental approach. In this paper, we describe scenarios where robotic systems interact with household objects by pushing them using robot arms while observing the scene with cameras, and which must incrementally learn, without external supervision, both the effect classes that emerge from these interactions as well as a discriminative model for predicting them from object properties. We formalize the scenario as a multi-view learning problem where data co-occur over two separate data views over time, and we present an online learning framework that uses a self-supervised form of learning vector quantization to build the discriminative model. In various experiments, we demonstrate the effectiveness of this approach in comparison with related supervised methods using data from experiments performed using two different robotic platforms.

Incorporating privileged information through metric learning

In some pattern analysis problems, there exists expert knowledge, in addition to the original data involved in the classification process. The vast majority of existing approaches simply ignore such auxiliary (privileged) knowledge. Recently a new paradigm-learning using privileged information-was introduced in the framework of SVM+. This approach is formulated for binary classification and, as typical for many kernel-based methods, can scale unfavorably with the number of training examples. While speeding up training methods and extensions of SVM+ to multiclass problems are possible, in this paper we present a more direct novel methodology for incorporating valuable privileged knowledge in the model construction phase, primarily formulated in the framework of generalized matrix learning vector quantization. This is done by changing the global metric in the input space, based on distance relations revealed by the privileged information. Hence, unlike in SVM+, any convenient classifier can be used after such metric modification, bringing more flexibility to the problem of incorporating privileged information during the training. Experiments demonstrate that the manipulation of an input space metric based on privileged data improves classification accuracy. Moreover, our methods can achieve competitive performance against the SVM+ formulations.

Analysis of flow cytometry data by matrix relevance learning vector quantization

Flow cytometry is a widely used technique for the analysis of cell populations in the study and diagnosis of human diseases. It yields large amounts of high-dimensional data, the analysis of which would clearly benefit from efficient computational approaches aiming at automated diagnosis and decision support. This article presents our analysis of flow cytometry data in the framework of the DREAM6/FlowCAP2 Molecular Classification of Acute Myeloid Leukemia (AML) Challenge, 2011. In the challenge, example data was provided for a set of 179 subjects, comprising healthy donors and 23 cases of AML. The participants were asked to provide predictions with respect to the condition of 180 patients in a test set. We extracted feature vectors from the data in terms of single marker statistics, including characteristic moments, median and interquartile range of the observed values. Subsequently, we applied Generalized Matrix Relevance Learning Vector Quantization (GMLVQ), a machine learning technique which extends standard LVQ by an adaptive distance measure. Our method achieved the best possible performance with respect to the diagnoses of test set patients. The extraction of features from the flow cytometry data is outlined in detail, the machine learning approach is discussed and classification results are presented. In addition, we illustrate how GMLVQ can provide deeper insight into the problem by allowing to infer the relevance of specific markers and features for the diagnosis.

Trainable COSFIRE filters for keypoint detection and pattern recognition

BACKGROUND

Keypoint detection is important for many computer vision applications. Existing methods suffer from insufficient selectivity regarding the shape properties of features and are vulnerable to contrast variations and to the presence of noise or texture.

METHODS

We propose a trainable filter which we call Combination Of Shifted FIlter REsponses (COSFIRE) and use for keypoint detection and pattern recognition. It is automatically configured to be selective for a local contour pattern specified by an example. The configuration comprises selecting given channels of a bank of Gabor filters and determining certain blur and shift parameters. A COSFIRE filter response is computed as the weighted geometric mean of the blurred and shifted responses of the selected Gabor filters. It shares similar properties with some shape-selective neurons in visual cortex, which provided inspiration for this work.

RESULTS

We demonstrate the effectiveness of the proposed filters in three applications: the detection of retinal vascular bifurcations (DRIVE dataset: 98.50 percent recall, 96.09 percent precision), the recognition of handwritten digits (MNIST dataset: 99.48 percent correct classification), and the detection and recognition of traffic signs in complex scenes (100 percent recall and precision).

CONCLUSIONS

The proposed COSFIRE filters are conceptually simple and easy to implement. They are versatile keypoint detectors and are highly effective in practical computer vision applications.