Dynamics of Visual Supervised Learning: A Statistical-Physics Approach

Perception ◽  
1996 ◽  
Vol 25 (1_suppl) ◽  
pp. 165-165
Author(s):  
A Unzicker ◽  
M Jüttner ◽  
I Rentschler
Keyword(s):  
Supervised Learning ◽  
Statistical Physics ◽  
Energy Gap ◽  
Feature Space ◽  
Mathematical Structure ◽  
Thermodynamic System ◽  
Classification Performance ◽  
Classification Error ◽  
Wide Range ◽  
Supervised Learning And Classification

We analysed human supervised learning and classification performance for compound Gabor gray-level patterns. We found that internal visual representations for supervised learning and classification may not be constructed in a smooth process of gradual development (Jüttner and Rentschler, 1996 Vision Research in press). Rather, it seemed that certain learning states (‘stereotypes’) recur that may be considered as ‘perceptual hypotheses’. Such effects have a transient character and cannot, therefore, be studied on the basis of cumulative learning data, which allow smoothing at the expense of temporal resolution. Thus, we analyse classification behaviour in terms of the evolution of a thermodynamic system, that is a system characterised by Gibbs statistics. Here it is assumed that a classification error occurs when a noise-influenced decision process passes an ‘energy gap’ related to the distance of signals in feature space. This approach has been extended to a wide range of distance-based models, originated by different fields, such as classical psychometrics, signal detection theory, technical pattern recognition, and connectionism. We made use of the finding that all these models can be related to a uniform mathematical structure (Unzicker et al, 1995 Perception24 Supplement, 95). The subjects' performance can then be described as a cooling process that reveals adaptive feature extraction during learning.

Electrocardiogram stream level correlated patterns as features to classify heartbeats for arrhythmia prediction

2020 ◽  
Vol 54 (5) ◽  
pp. 685-701
Author(s):  
Fuad Ali Mohammed Al-Yarimi ◽  
Nabil Mohammed Ali Munassar ◽  
Fahd N. Al-Wesabi
Keyword(s):  
Supervised Learning ◽  
Computation Time ◽  
Classification Error ◽  
Clinical Practices ◽  
Content Type ◽  
A Value ◽  
Correlated Patterns ◽  
Constructive Models ◽  
Minimum Classification Error ◽  
Supervised Learning And Classification

PurposeDigital computing and machine learning-driven predictive analysis in the diagnosis of non-communicable diseases are gaining significance. Globally many research studies are focusing on developing comprehensive models for such detection. Categorically in the proposed diagnosis for arrhythmia, which is a critical diagnosis to prevent cardiac-related deaths, any constructive models can be a value proposition. In this study, the focus is on developing a holistic system that predicts the scope of arrhythmia from the given electrocardiogram report. The proposed method is using the sequential patterns of the electrocardiogram elements as features.Design/methodology/approachConsidering the decision accuracy of the contemporary classification methods, which is not adequate to use in clinical practices, this manuscript coined a new dimension of features to perform supervised learning and classification using the AdaBoost classifier. The proposed method has titled “Electrocardiogram stream level correlated patterns as features (ESCPFs),” which takes electrocardiograms (ECGs) signal streams as input records to perform supervised learning-based classification to detect the arrhythmia scope in given ECG record.FindingsFrom the results and comparative reports generated for the study, it is evident that the model is performing with higher accuracy compared to some of the earlier models. However, focusing on the emerging solutions and technologies, if the accuracy factors for the model can be improved, it can lead to compelling predictions and accurate outcome from the process.Originality/valueThe authors represent complete automatic and rapid arrhythmia as classifier, which could be applied online and examine long ECG records sequence efficiently. By releasing the needs for extraction of features, the authors project an application based on raw signals, one result to heart rates date, whose objective is to lessen computation time when attaining minimum classification error outcomes.

scholarly journals Supervised Learning With First-to-Spike Decoding in Multilayer Spiking Neural Networks

2021 ◽  
Vol 15 ◽  
Author(s):  
Brian Gardner ◽  
André Grüning
Keyword(s):  
Neural Networks ◽  
Supervised Learning ◽  
Experimental Studies ◽  
Learning Rule ◽  
Image Data ◽  
Classification Performance ◽  
Spiking Neural Networks ◽  
Classification Problems ◽  
Sensory Stimuli ◽  
Wide Range

Experimental studies support the notion of spike-based neuronal information processing in the brain, with neural circuits exhibiting a wide range of temporally-based coding strategies to rapidly and efficiently represent sensory stimuli. Accordingly, it would be desirable to apply spike-based computation to tackling real-world challenges, and in particular transferring such theory to neuromorphic systems for low-power embedded applications. Motivated by this, we propose a new supervised learning method that can train multilayer spiking neural networks to solve classification problems based on a rapid, first-to-spike decoding strategy. The proposed learning rule supports multiple spikes fired by stochastic hidden neurons, and yet is stable by relying on first-spike responses generated by a deterministic output layer. In addition to this, we also explore several distinct, spike-based encoding strategies in order to form compact representations of presented input data. We demonstrate the classification performance of the learning rule as applied to several benchmark datasets, including MNIST. The learning rule is capable of generalizing from the data, and is successful even when used with constrained network architectures containing few input and hidden layer neurons. Furthermore, we highlight a novel encoding strategy, termed “scanline encoding,” that can transform image data into compact spatiotemporal patterns for subsequent network processing. Designing constrained, but optimized, network structures and performing input dimensionality reduction has strong implications for neuromorphic applications.

scholarly journals Laplacian Support Vector Machine for Vibration-Based Robotic Terrain Classification

Electronics ◽  
2020 ◽  
Vol 9 (3) ◽  
pp. 513 ◽  
Author(s):  
Wenlei Shi ◽  
Zerui Li ◽  
Wenjun Lv ◽  
Yuping Wu ◽  
Ji Chang ◽  
...  
Keyword(s):  
Supervised Learning ◽  
Classification Accuracy ◽  
Temporal Correlation ◽  
Feature Space ◽  
Classification Performance ◽  
Support Vector ◽  
Terrain Classification ◽  
Potential Threat ◽  
Temporal Space ◽  
Unlabelled Data

The achievement of robot autonomy has environmental perception as a prerequisite. The hazards rendered from uneven, soft and slippery terrains, which are generally named non-geometric hazards, are another potential threat reducing the traversing efficient, and therefore receiving more and more attention from the robotics community. In the paper, the vibration-based terrain classification (VTC) is investigated by taking a very practical issue, i.e., lack of labels, into consideration. According to the intrinsic temporal correlation existing in the sampled terrain sequence, a modified Laplacian SVM is proposed to utilise the unlabelled data to improve the classification performance. To the best of our knowledge, this is the first paper studying semi-supervised learning problem in robotic terrain classification. The experiment demonstrates that: (1) supervised learning (SVM) achieves a relatively low classification accuracy if given insufficient labels; (2) feature-space homogeneity based semi-supervised learning (traditional Laplacian SVM) cannot improve supervised learning’s accuracy, and even makes it worse; (3) feature- and temporal-space based semi-supervised learning (modified Laplacian SVM), which is proposed in the paper, could increase the classification accuracy very significantly.

Feature space approximation for kernel-based supervised learning

2021 ◽  
Vol 221 ◽  
pp. 106935
Author(s):  
Patrick Gelß ◽  
Stefan Klus ◽  
Ingmar Schuster ◽  
Christof Schütte
Keyword(s):  
Supervised Learning ◽  
Feature Space

scholarly journals The Band-Gap Studies of Short-Period CdO/MgO Superlattices

2021 ◽  
Vol 16 (1) ◽  
Author(s):  
Ewa Przeździecka ◽  
P. Strąk ◽  
A. Wierzbicka ◽  
A. Adhikari ◽  
A. Lysak ◽  
...  
Keyword(s):  
Band Gap ◽  
Layer Thickness ◽  
Energy Gap ◽  
Band Gaps ◽  
Short Period ◽  
Wide Range ◽  
Salt Structure ◽  
Short Period Superlattices ◽  
Direct Band ◽  
Sublayer Thickness

AbstractTrends in the behavior of band gaps in short-period superlattices (SLs) composed of CdO and MgO layers were analyzed experimentally and theoretically for several thicknesses of CdO sublayers. The optical properties of the SLs were investigated by means of transmittance measurements at room temperature in the wavelength range 200–700 nm. The direct band gap of {CdO/MgO} SLs were tuned from 2.6 to 6 eV by varying the thickness of CdO from 1 to 12 monolayers while maintaining the same MgO layer thickness of 4 monolayers. Obtained values of direct and indirect band gaps are higher than those theoretically calculated by an ab initio method, but follow the same trend. X-ray measurements confirmed the presence of a rock salt structure in the SLs. Two oriented structures (111 and 100) grown on c- and r-oriented sapphire substrates were obtained. The measured lattice parameters increase with CdO layer thickness, and the experimental data are in agreement with the calculated results. This new kind of SL structure may be suitable for use in visible, UV and deep UV optoelectronics, especially because the energy gap can be precisely controlled over a wide range by modulating the sublayer thickness in the superlattices.

scholarly journals Student Academic Performance Prediction using Supervised Learning Techniques

2019 ◽  
Vol 14 (14) ◽  
pp. 92 ◽  
Author(s):  
Muhammad Imran ◽  
Shahzad Latif ◽  
Danish Mehmood ◽  
Muhammad Saqlain Shah
Keyword(s):  
Data Mining ◽  
Supervised Learning ◽  
Student Performance ◽  
Performance Prediction ◽  
Class Imbalance ◽  
Ensemble Methods ◽  
Fine Tuning ◽  
Classification Error ◽  
Decision Tree Classifier ◽  
Tree Classifier

Automatic Student performance prediction is a crucial job due to the large volume of data in educational databases. This job is being addressed by educational data mining (EDM). EDM develop methods for discovering data that is derived from educational environment. These methods are used for understanding student and their learning environment. The educational institutions are often curious that how many students will be pass/fail for necessary arrangements. In previous studies, it has been observed that many researchers have intension on the selection of appropriate algorithm for just classification and ignores the solutions of the problems which comes during data mining phases such as data high dimensionality ,class imbalance and classification error etc. Such types of problems reduced the accuracy of the model. Several well-known classification algorithms are applied in this domain but this paper proposed a student performance prediction model based on supervised learning decision tree classifier. In addition, an ensemble method is applied to improve the performance of the classifier. Ensemble methods approach is designed to solve classification, predictions problems. This study proves the importance of data preprocessing and algorithms fine-tuning tasks to resolve the data quality issues. The experimental dataset used in this work belongs to Alentejo region of Portugal which is obtained from UCI Machine Learning Repository. Three supervised learning algorithms (J48, NNge and MLP) are employed in this study for experimental purposes. The results showed that J48 achieved highest accuracy 95.78% among others.

Semi-Supervised Classification and its Application to Filtering IDS False Positives

2013 ◽  
Vol 427-429 ◽  
pp. 2309-2312
Author(s):  
Hai Bin Mei ◽  
Ming Hua Zhang
Keyword(s):  
Supervised Learning ◽  
Supervised Classification ◽  
Classification Performance ◽  
False Positives ◽  
Training Data ◽  
Classification Model ◽  
Classification Technique

Alert classifiers built with the supervised classification technique require large amounts of labeled training alerts. Preparing for such training data is very difficult and expensive. Thus accuracy and feasibility of current classifiers are greatly restricted. This paper employs semi-supervised learning to build alert classification model to reduce the number of needed labeled training alerts. Alert context properties are also introduced to improve the classification performance. Experiments have demonstrated the accuracy and feasibility of our approach.

scholarly journals Structure and function relationships of proteins based on polar profile: a review.

2016 ◽  
Vol 63 (2) ◽  
Author(s):  
Carlos Polanco ◽  
Thomas Buhse ◽  
Vladimir N Uversky
Keyword(s):  
Mathematical Structure ◽  
Quantitative Structure Activity Relationship ◽  
Multidrug Resistant ◽  
Training Data ◽  
Main Function ◽  
Pharmaceutical Drugs ◽  
Linear Sequence ◽  
Wide Range ◽  
High Level ◽  
And Function

Proteins in the post-genome era impose diverse research challenges, the main are the understanding of their structure-function mechanism, and the growing need for new pharmaceutical drugs, particularly antibiotics that help clinicians treat the ever- increasing number of Multidrug-Resistant Organisms (MDROs). Although, there is a wide range of mathematical-computational algorithms to satisfy the demand, among them the Quantitative Structure-Activity Relationship algorithms that have shown better performance using a characteristic training data of the property searched; their performance has stagnated regardless of the number of metrics they evaluate and their complexity. This article reviews the characteristics of these metrics, and the need to reconsider the mathematical structure that expresses them, directing their design to a more comprehensive algebraic structure. It also shows how the main function of a protein can be determined by measuring the polarity of its linear sequence, with a high level of accuracy, and how such exhaustive metric stands as a "fingerprint" that can be applied to scan the protein regions to obtain new pharmaceutical drugs, and thus to establish how the singularities led to the specialization of the protein groups known today.

scholarly journals Markovian analysis of cervical cell images.

1976 ◽  
Vol 24 (1) ◽  
pp. 138-144 ◽  
Author(s):  
N J Pressman
Keyword(s):  
Transition Probabilities ◽  
Classification Performance ◽  
Error Rates ◽  
Classification Error ◽  
Correct Classification ◽  
Cell Nuclei ◽  
Step Size ◽  
Cervical Cell ◽  
Texture Parameters ◽  
Markovian Analysis

Markovian analysis is a method to measure optical texture based on gray-level transition probabilities in digitized images. Experiments are described that investigate that classification performance of parameters generated by Markovian analysis. Results using Markov texture parameters show that the selection of a Markov step size strongly affects classification error rates and the number of parameters required to achieve the maximum correct classification rates. Markov texture parameters are shown to achieve high rates of correct classification in discriminating images of normal from abnormal cervical cell nuclei.

40 Years Trajectory of Amorphous Semiconductor Research

2000 ◽  
Vol 609 ◽  
Author(s):  
Yoshihiro Hamakawa
Keyword(s):  
Thin Film ◽  
Solar Cells ◽  
Energy Gap ◽  
Early Period ◽  
Basic Research ◽  
Amorphous Semiconductor ◽  
Film Deposition ◽  
Terrestrial Environment ◽  
Heterojunction Solar Cells ◽  
Wide Range

ABSTRACTA review is given on a research trajectory of amorphous and microcrystalline semiconductors and their device applications proceeded since 1970. A brief explanation on the motivation to start amorphous semiconductor research is given to produce a new kind of synthetic semiconductor having continuous energy gap controllability with valency electron controllability through our experience of modulation spectroscopy in semiconductors.The first material we have challenged is Si-As-Te chalcogenide semiconductor which has a very wide vitreous region in Gibb's Triangle. A series of systematic experiments has been carried out in the terrestrial environment since 1971, and also within the TT-500A rocket experiment in 1980, and the Spacelab. J experiments FMPT (First Material Processing Test) project in 1992. The second material is hydrogenated amorphous silicon (a-Si:H) and its alloys started in 1976 just after the Garmisch Partenkirchen ICALS-6. With some basic research on the a-Si:H film deposition technology and film quality improvement, our continuous effort to improve the efficiency bore the tandem type solar cells in 1979, and also new products of a-SiC:H and a-SiGe:H in the early period of 1980s are described. These innovative device structures and materials have bloomed in the middle of 1980s in R & D phase such as a-SiC/a-Si heterojunction solar cells, a-Si/a-SiGe and also a-Si/poly-Si tandem type solar cells, and industrialized in recent few years. New kind of trials on full-color thin film light emitting devices has also been recently initiated with wide range of band gap controllability of a-SiC:H.The third material is microcrystalline silicon (µc-Si) and their alloys which gathers a tremendous R & D effort as a promised candidate for the bottom cell of the a-Si/µc-Si tandem solar cells aimed for the all-round plasma CVD process for the next age thin film photovoltaic devices. In the final part of presentation, a brief discussion will be given on a technological evolution from “bulk crystalline age” to “multilayered thin film age” in the semiconductor optoelectronics toward 21 century.

Sign in / Sign up

Export Citation Format

Share Document