Search results for "incremental learning"
showing 10 items of 13 documents
Panel Summary Perceptual Learning and Discovering
1994
The problem of learning and discovering in perception is addressed and discussed with particular reference to present machine learning paradigms. These paradigms are briefly introduced by S. Gaglio. The subsymbolic approach is addressed by S. Nolfi, and the role of symbolic learning is analysed by F. Esposito. Many of the open problems, that are evidentiated in the course of the panel, show how this is an important field of research that still needs a lot of investigation. In particular, as a result of the whole discussion, it seems that a suitable integration of different approaches must be accurately investigated. It is observed, in fact, that the weakness of the most part of the existing…
An integrated fuzzy cells-classifier
2007
This paper introduces a genetic algorithm able to combine different classifiers based on different distance functions. The use of a genetic algorithm is motivated by the fact that the combination phase is based on the optimization of a vote strategy. The method has been applied to the classification of four types of biological cells, results show an improvement of the recognition rate using the genetic algorithm combination strategy compared with the recognition rate of each single classifier.
Incremental Gaussian Discriminant Analysis based on Graybill and Deal weighted combination of estimators for brain tumour diagnosis
2011
In the last decade, machine learning (ML) techniques have been used for developing classifiers for automatic brain tumour diagnosis. However, the development of these ML models rely on a unique training set and learning stops once this set has been processed. Training these classifiers requires a representative amount of data, but the gathering, preprocess, and validation of samples is expensive and time-consuming. Therefore, for a classical, non-incremental approach to ML, it is necessary to wait long enough to collect all the required data. In contrast, an incremental learning approach may allow us to build an initial classifier with a smaller number of samples and update it incrementally…
A Random Extension for Discriminative Dimensionality Reduction and Metric Learning
2009
A recently proposed metric learning algorithm which enforces the optimal discrimination of the different classes is extended and empirically assessed using different kinds of publicly available data. The optimization problem is posed in terms of landmark points and then, a stochastic approach is followed in order to bypass some of the problems of the original algorithm. According to the results, both computational burden and generalization ability are improved while absolute performance results remain almost unchanged.
OnMLM: An Online Formulation for the Minimal Learning Machine
2019
Minimal Learning Machine (MLM) is a nonlinear learning algorithm designed to work on both classification and regression tasks. In its original formulation, MLM builds a linear mapping between distance matrices in the input and output spaces using the Ordinary Least Squares (OLS) algorithm. Although the OLS algorithm is a very efficient choice, when it comes to applications in big data and streams of data, online learning is more scalable and thus applicable. In that regard, our objective of this work is to propose an online version of the MLM. The Online Minimal Learning Machine (OnMLM), a new MLM-based formulation capable of online and incremental learning. The achievements of OnMLM in our…
Motion analysis using the novelty filter
1991
Abstract An original approach to the motion analysis, based on the novelty filter, is proposed. The novelty filter stresses the novelties occurring in a pattern representing an image of the scene under consideration with respect to patterns representing previous images of the same scene, so that visual information about the motion of the objects is obtained. The novelty filter may be implemented by a neural network architecture, taking advantage of the capabilities of massive parallelism, adaptive learning and noise robustness. The novelty filter may learn the entire trajectory of an object, through an incremental learning of a sequence of images capturing the scene, thus emphasizing if the…
A Grid Enabled Parallel Hybrid Genetic Algorithm for SPN
2004
This paper presents a combination of a parallel Genetic Algorithm (GA) and a local search methodology for the Steiner Problem in Networks (SPN). Several previous papers have proposed the adoption of GAs and others metaheuristics to solve the SPN demonstrating the validity of their approaches. This work differs from them for two main reasons: the dimension and the features of the networks adopted in the experiments and the aim from which it has been originated. The reason that aimed this work was namely to assess deterministic and computationally inexpensive algorithms which can be used in practical engineering applications, such as the multicast transmission in the Internet. The large dimen…
A novel identification method for generalized T-S fuzzy systems
2012
Published version of an article from the journal: Mathematical Problems in Engineering. Also available from the publisher:http://dx.doi.org/10.1155/2012/893807 In order to approximate any nonlinear system, not just affine nonlinear systems, generalized T-S fuzzy systems, where the control variables and the state variables, are all premise variables are introduced in the paper. Firstly, fuzzy spaces and rules were determined by using ant colony algorithm. Secondly, the state-space model parameters are identified by using genetic algorithm. The simulation results show the effectiveness of the proposed algorithm
An Algorithm for Discretization of Real Value Attributes Based on Interval Similarity
2013
Discretization algorithm for real value attributes is of very important uses in many areas such as intelligence and machine learning. The algorithms related to Chi2 algorithm (includes modified Chi2 algorithm and extended Chi2 algorithm) are famous discretization algorithm exploiting the technique of probability and statistics. In this paper the algorithms are analyzed, and their drawback is pointed. Based on the analysis a new modified algorithm based on interval similarity is proposed. The new algorithm defines an interval similarity function which is regarded as a new merging standard in the process of discretization. At the same time, two important parameters (condition parameterαand ti…
CN2-R: Faster CN2 with randomly generated complexes
2011
Among the rule induction algorithms, the classic CN2 is still one of the most popular ones; a great amount of enhancements and improvements to it is to witness this. Despite the growing computing capacities since the algorithm was proposed, one of the main issues is resource demand. The proposed modification, CN2-R, substitutes the star concept of the original algorithm with a technique of randomly generated complexes in order to substantially improve on running times without significant loss in accuracy.