Enhance Classification Accuracy Research Articles

Enhancing Classification Accuracy of Transhumeral Prosthesis: A Hybrid sEMG and fNIRS Approach

Limited non-invasive transhumeral prosthesis control exists due to the absence of signal sources on amputee residual muscles. This paper introduces a hybrid brain-machine interface (hBMI) that integrates surface electromyography (sEMG) and functional near-infrared spectroscopy (fNIRS) signals to overcome the limits of an existing myoelectric upper-limb prosthesis. This hybridization aims to improve classification accuracy (CA) to escalate arm movements' control performance for individuals who have a transhumeral amputation. To evaluate the effectiveness of this hBMI, fifteen healthy and three transhumeral amputee subjects for six arm motions were participating in the experiment. Myo armband was used to acquire sEMG signals corresponding to four arm motions: elbow extension, elbow flexion, wrist pronation, and wrist supination. Whereas, fNIRS brain imaging modality was used to monitor cortical hemodynamics response from the prefrontal cortex region for two hand motions: hand open and hand close. The average accuracy of 94.6 % and 74% was achieved for elbow and wrist motions by sEMG for healthy and amputated subjects, respectively. Simultaneously, the fNIRS modality showed an average accuracy of 96.9% and 94.5% for hand motions of healthy and amputated subjects. This study demonstrates the feasibility of hybridizing sEMG and fNIRS signals to improve the CA for transhumeral amputees, improving the control performances of multifunctional upper-limb prostheses.

Enhancement of classification accuracy of multi-spectral satellites’ images using Laplacian pyramids

Processed satellites’ images are not widely used in remote sensing classification due to the changes in spectral properties that may confuse the classifiers. In pixel-based classification there is a certain debate concerning with the boundary pixels. Most of miss-classified pixels are boundary pixels due to the sudden change in the spectral properties of the contacted objects. This research work is an investigation to study the proper enhancement in classification accuracy that may occur if the Laplacian pyramids are used in classification. The reference map is prepared to study the performance of the proposed system. The Laplacian image is constructed for each band of the satellite image. Then the classification is carried out for both the Laplacian image pyramid and the original satellite image using competitive learning neural networks (CLNN) method. The evaluation is carried out by comparing the classified Laplacian image with the classified original image. A statistical test is carried out to study the significance of using the classified Laplacian image in classification.

Demonstration of the applicability of visible and near-infrared spatially resolved spectroscopy for rapid and nondestructive wood classification

Abstract Although visible and near-infrared (Vis-NIR) spectroscopy can rapidly and nondestructively identify wood species, the conventional spectrometer approach relies on the aggregate light absorption due to the chemical composition of wood and light scattering originating from the physical structure of wood. Hence, much of the work in this area is still limited to further spectral pretreatments, such as baseline correction and standard normal variate to reduce the light scattering effects. However, it should be emphasized that the light scattering rather than absorption in wood is dominant, and this must be effectively utilized to achieve highly accurate and robust wood classification. Here a novel method based on spatially resolved diffuse reflectance (wavelength range: 600–1000 nm) was demonstrated to classify 15 kinds of wood. A portable Vis-NIR spectral measurement system was designed according to previous simulations and experimental results. To simplify spectral data analysis (i.e., against overfitting), support vector machine (SVM) model was constructed for wood sample classification using principal component analysis (PCA) scores. The classification accuracies of 98.6% for five-fold cross-validation and 91.2% for test set validation were achieved. This study offers enhanced classification accuracy and robustness over other conventional nondestructive approaches for such various kinds of wood and sheds light on utilizing visible and short-wave NIR light scattering for wood classification.

Deep learning for liver tumour classification: enhanced loss function

Background and Aim: deep learning has not been successfully implemented in liver tumour feature extraction and classification using computer-aided diagnosis. This study aims to enhance classification accuracy and improves the processing time to better differentiate tumour types. Methodology: This study proposed a hybrid model, which combines the regularization function with the current loss function for the support vector machine (SVM) classifier. Regularization function is used for prioritizing image classes before feeding it to the linear mapping. The proposed model consists of the region growing algorithm to get the region-of-interest (ROI), and Weiner filtering algorithm for image enhancement and noise removal. The gray level co-occurrence matrix (GLCM) was performed to extract the feature from the image. The extracted feature then fed to SVM classifier using selected feature vectors to classify the affected region and neglecting the unwanted areas. Results: classification accuracy was calculated using probability score, and the processing time was calculated based on the total execution time. The proposed system was able to achieve an average classification accuracy of 98.9%, which is about 2–3% higher than the current system. The results showed that 12 ms reduced the processing time on average. Conclusion: The proposed system focused on improving feature extraction and classification for different types of tumours from the MRI images. The study solved the problem in linear mapping of support vector machine and enhanced the classification accuracy and the processing time of early diagnosis of three different types of tumours in liver MRI images.

Enhancing classification accuracy of fNIRS-BCI using features acquired from vector-based phase analysis

Objective. In this paper, a novel methodology for feature extraction to enhance classification accuracy of functional near-infrared spectroscopy (fNIRS)-based two-class and three-class brain–computer interface (BCI) is presented. Approach. Novel features are extracted using vector-based phase analysis method. Changes in oxygenated and de-oxygenated ) haemoglobin are used to calculate four novel features: change in cerebral blood volume (), change in cerebral oxygen exchange (), vector magnitude (|L|) and angle (k). is the sum and is difference of and , whereas |L| is magnitude and k is angle of vector. fNIRS signals of seven healthy subjects, corresponding to left-hand index finger tapping (LFT), right-hand index finger tapping (RFT) and rest are acquired from motor cortex using multi-channel continuous-wave imaging system. After removing physiological and instrumental noises from the acquired signals, the four novel features are calculated. For validation, conventional temporal, spatial and spatiotemporal features; mean, peak, slope, variance, kurtosis and skewness are also calculated using and. All possible two-feature and three-feature combinations of the novel and conventional features are then used to classify two-class (LFT vs RFT) and three-class (LFT vs RFT vs rest) fNIRS-BCI using linear discriminant analysis. Main results. Results demonstrate that combination of four novel features yields significantly higher average classification accuracies of 98.7 ± 1.0% and 85.4 ± 1.4% as compared to 68.7 ± 6.9% and 53.6 ± 10.6% using conventional features for two-class and three-class problem, respectively. Validation of proposed method on an open access database containing RFT, LFT and dominant side foot tapping tasks for 30 subjects also shows improvement in average classification accuracies for two-class and three-class fNIRS-BCIs. Significance. This study provides a step forward in improving the classification accuracies of state-of-the-art fNIRS-BCIs by showing significant improvement in classification accuracies of two-class and three-class fNIRS-BCIs using novel features extracted by vector-based phase analysis.

GeFeS: A generalized wrapper feature selection approach for optimizing classification performance

In this paper, we propose a generalized wrapper-based feature selection, called GeFeS, which is based on a parallel new intelligent genetic algorithm (GA). The proposed GeFeS works properly under different numerical dataset dimensions and sizes, carefully tries to avoid overfitting and significantly enhances classification accuracy. To make the GA more accurate, robust and intelligent, we have proposed a new operator for features weighting, improved the mutation and crossover operators, and integrated nested cross-validation into the GA process to properly validate the learning model. The k-nearest neighbor (kNN) classifier is utilized to evaluate the goodness of selected features. We have evaluated the efficiency of GeFeS on various datasets selected from the UCI machine learning repository. The performance is compared with state-of-the-art classification and feature selection methods. The results demonstrate that GeFeS can significantly generalize the proposed multi-population intelligent genetic algorithm under different sizes of two-class and multi-class datasets. We have achieved the average classification accuracy of 95.83%, 97.62%, 99.02%, 98.51%, and 94.28% while reducing the number of features from 56 to 28, 34 to 18, 279 to 135, 30 to 16, and 19 to 9 under lung cancer, dermatology, arrhythmia, WDBC, and hepatitis, respectively.

TIME SERIES LAND COVER CLASSIFICATION BASED ON SEMI-SUPERVISED CONVOLUTIONAL LONG SHORT-TERM MEMORY NEURAL NETWORKS

Abstract. Time series imagery containing high-dimensional temporal features are conducive to improving classification accuracy. With the plenty accumulation of historical images, the inclusion of time series data becomes available to utilize, but it is difficult to avoid missing values caused by cloud cover. Meanwhile, seeking a large amount of training labels for long time series also makes data collection troublesome. In this study, we proposed a semi-supervised convolutional long short-term memory neural network (Semi-LSTM) in long time series which achieves an accurate and automated land cover classification with a small proportion of labels. Three main contributions of this work are summarized as follows: i) the proposed method achieve an excellent classification via a small group of labels in long time series data, and reducing dependence of training labels; ii) it is a robust algorithm in accuracy for the influence of noise, and reduces the requirements of sequential data for cloudless and lossless images; and iii) it makes full advantage of spectral-spatial-temporal features, especially expanding time context information to enhance classification accuracy. Finally, the proposed network is validated on time series imagery from Landsat 8. All quantitative analyses and evaluation indicators of the experimental results demonstrate competitive performance in the suggested modes.

GMDH-Based Outlier Detection Model in Classification Problems

In many practical classification problems, datasets would have a portion of outliers, which could greatly affect the performance of the constructed models. In order to address this issue, we apply the group method of data handing (GMDH) neural network in outlier detection. This study builds a GMDH-based outlier detection (GOD) model. This model first implements feature selection in the training set L using GMDH neural network. Then a new training set L′ can be obtained by mapping the selected key feature subset. Next, a linear regression model can be constructed in the set L′ by ordinary least squares estimation. Further, it eliminates a sample from the set L′ randomly every time, and then rebuilds a linear regression model. Finally, outlier detection is realized by calculating Cook’s distance for each sample. Four different customer classification datasets are used to conduct experiments. Results show that GOD model can effectively eliminate outliers, and compared with the five existing outlier detection models, it generally performs significantly better. This indicates that eliminating outliers can effectively enhance classification accuracy of the trained classification model.

Improved Land Cover Mapping Using Landsat 8 Thermal Imagery

Detection of land cover (LC) changes allows policymakers to recognize the complexities of environmental modification and change to achieve sustainability of economic growth. As a result, recognition of LC features has appeared as an essential research dimension and, consequently, an appropriate and reliable methodology for classifying LC is occasionally required. In this research, Landsat 8 satellite data captured by Operational Land Imager (OLI) and Thermal Infrared Scanner (TIRS) were utilized for the LC classification using the Support Vector Machine (SVM) classifier algorithm. The aim of the study is to enhance classification accuracy by integrating the use of data from satellite thermal and spectral imaging. Land Surface Temperature (LST) is sensitive to the soil surface characteristics, therefore, it may be used to gather LC feature information. The classification accuracy was designed to enhance the integration of thermal information from Landsat 8’s thermal band TIRS and Landsat 8 OLI’s spectral data. In this study, Advanced Thermal Integrated Vegetation Index (ATLIVI) and Thermal Integrated Vegetation Index (TLIVI) established and revealed fairly strong correlations with the related surface temperature (Ts) by R2=0,7 and 0,65 respectively. The relationship between Ts and the other vegetation indices based on the empirical parameterization demonstrate that these two indices showed an improvement of almost 6% in the overall accuracy of the LC classification results compared to the Landsat 8 Standard False Colour Composite image as an input data using SVM algorithm.

Combining single photon and multispectral airborne laser scanning for land cover classification

Feasibility of using single photon lidar (SPL) data for land cover classification was studied for the first time. Data from the Espoonlahti suburban study area in Finland were acquired with the SPL100 sensor from a flying altitude of 3600 m above ground level and classified into six classes using an object-based, random forest classification approach. The overall accuracy was 66% compared with a set of reference points. The results were good for buildings and trees, but there was significant confusion between ground-level land cover classes. The possibility to enhance classification accuracy by combining SPL data with multispectral intensity information from airborne laser scanning (ALS) was also tested and led to an overall accuracy of 86%. The multispectral intensity image was obtained by combining intensity data from two channels of the Optech Titan multispectral ALS sensor (infrared and near-infrared) with the SPL intensity (green). To allow for comparisons between SPL and conventional linear-mode ALS, single-channel data from the Optech Titan (near-infrared) were also applied for the same classification task. In this case, the overall accuracy was 83%. An important factor affecting the lower quality of the SPL results was the rough appearance of the intensity images created from the SPL data. Despite high point density, small features such as narrow roads were often difficult to distinguish in the intensity images. This relates to the SPL technique and preprocessing operations applied to the data. Considering future use of new ALS datasets, the different benefits of different systems need to be taken into account. A clear benefit of the SPL technique is its high efficiency in data acquisition. SPL is seemingly fit for deriving high objects that are distinguishable by using geometric rather than intensity information. The classification accuracy of buildings and trees was comparable in each classification test, although the flying altitude of the SPL100 was five times that of the Optech Titan. We propose that raw echo waveforms would be provided together with point cloud data to allow users to create their own intensity and echo width features usable for classification purposes. Combining multispectral intensity with SPL, in reference to multispectral single photon lidar, could result in a technique that is effective for mapping large areas and will also allow for automated classification of ground-level objects in a more reliable manner.

A comparison of high-throughput plasma NMR protocols for comparative untargeted metabolomics

IntroductionWhen analyzing the human plasma metabolome with Nuclear Magnetic Resonance (NMR) spectroscopy, the Carr–Purcell–Meiboom–Gill (CPMG) experiment is commonly employed for large studies. However, this process can lead to compromised statistical analyses due to residual macromolecule signals. In addition, the utilization of Trimethylsilylpropanoic acid (TSP) as an internal standard often leads to quantification issues, and binning, as a spectral summarization step, can result in features not clearly assignable to metabolites.ObjectivesOur aim was to establish a new complete protocol for large plasma cohorts collected with the purpose of describing the comparative metabolic profile of groups of samples.MethodsWe compared the conventional CPMG approach to a novel procedure that involves diffusion NMR, using the Longitudinal Eddy-Current Delay (LED) experiment, maleic acid (MA) as the quantification reference and peak picking for spectral reduction. This comparison was carried out using the ultrafiltration method as a gold standard in a simple sample classification experiment, with Partial Least Squares–Discriminant Analysis (PLS-DA) and the resulting metabolic signatures for multivariate data analysis. In addition, the quantification capabilities of the method were evaluated.ResultsWe found that the LED method applied was able to detect more metabolites than CPMG and suppress macromolecule signals more efficiently. The complete protocol was able to yield PLS-DA models with enhanced classification accuracy as well as a more reliable set of important features than the conventional CPMG approach. Assessment of the quantitative capabilities of the method resulted in good linearity, recovery and agreement with an established amino acid assay for the majority of the metabolites tested. Regarding repeatability, ~ 85% of all peaks had an adequately low coefficient of variation (< 30%) in replicate samples.ConclusionOverall, our comparison yielded a high-throughput untargeted plasma NMR protocol for optimized data acquisition and processing that is expected to be a valuable contribution in the field of metabolic biomarker discovery.

Patchy Image Structure Classification Using Multi-Orientation Region Transform

Exterior contour and interior structure are both vital features for classifying objects. However, most of the existing methods consider exterior contour feature and internal structure feature separately, and thus fail to function when classifying patchy image structures that have similar contours and flexible structures. To address above limitations, this paper proposes a novel Multi-Orientation Region Transform (MORT), which can effectively characterize both contour and structure features simultaneously, for patchy image structure classification. MORT is performed over multiple orientation regions at multiple scales to effectively integrate patchy features, and thus enables a better description of the shape in a coarse-to-fine manner. Moreover, the proposed MORT can be extended to combine with the deep convolutional neural network techniques, for further enhancement of classification accuracy. Very encouraging experimental results on the challenging ultra-fine-grained cultivar recognition task, insect wing recognition task, and large variation butterfly recognition task are obtained, which demonstrate the effectiveness and superiority of the proposed MORT over the state-of-the-art methods in classifying patchy image structures. Our code and three patchy image structure datasets are available at: https://github.com/XiaohanYu-GU/MReT2019.

Human Identification Using Selected Features From Finger Geometric Profiles

A finger biometric system at an unconstrained environment is presented in this paper. A technique for hand image normalization is implemented at the preprocessing stage that decomposes the main hand contour into finger-level shape representation. This normalization technique follows subtraction of transformed binary image from binary hand contour image to generate the left-side of finger profiles (LSFPs). Then, XOR is applied to LSFP image and hand contour image to produce the right side of finger profiles. During feature extraction, initially, 30 geometric features are computed from every normalized finger. The rank-based forward–backward greedy algorithm is followed to select relevant features and to enhance classification accuracy. Two different subsets of features containing 9 and 12 discriminative features per finger are selected for two separate experimentations those use the ${k}$ -nearest neighbor and the random forest (RF) for classification on the Bosphorus hand database. The experiments with the selected features of four fingers except the thumb have obtained improved performances compared to features extracted from five fingers and also other existing methods evaluated on the Bosphorus database. The best identification accuracies of 96.56% and 95.92% using the RF classifier have been achieved for the right- and left-hand images of 638 subjects, respectively. An equal error rate of 0.078 is obtained for both types of the hand images.

Synthetic-Range-Profile-Based Training Library Construction for Ship Target Recognition Purposes of Scanning Radar Systems

The quantity and quality of range profiles in a training library play a significant role in target recognition. Classifier structures should be built up by experiencing a sufficient amount of range profiles in order to enhance classification accuracy. On the other hand, there might be insufficient measuremental range profiles especially for the targets rarely cruising. Therefore, training libraries should be enriched for these types of targets. In this article, the suitability and efficiency of creating a training library with synthetic range profiles are investigated for maritime target recognition purposes of scanning radar systems. In this context, synthetic range profiles are generated for six different ship targets, and the compatibility of these profiles with measuremental ones has been analyzed from different points of view. Moreover, a novel approach on generating synthetic range profiles is introduced. In the sense of compatibility studies, first, correlation analyses have been performed directly between range profiles and then between spatial- and velocity-based features. It is seen that the proposed approach has the ability to generate more compatible range profiles than the conventional methodology. In addition to similarity analyses, classification examinations have been performed for different feature use cases by employing synthetic profiles in the training set and the measuremental ones in the test set. At this step, a k-nearest-neighbor-based classifier that benefits by the discrimination ability of length- and velocity-based features is proposed to obtain higher classification rates. The results of the classification examinations show that synthetic profiles could be used either to construct an offline training library from scratch or to enrich a nonhomogenous and lacking library for ship target recognition.

Inter-Seasonal Time Series Imagery Enhances Classification Accuracy of Grazing Resource and Land Degradation Maps in a Savanna Ecosystem

In savannas, mapping grazing resources and indicators of land degradation is important for assessing ecosystem conditions and informing grazing and land management decisions. We investigated the effects of classifiers and used time series imagery—images acquired within and across seasons—on the accuracy of plant species maps. The study site was a grazed savanna in southern Kenya. We used Sentinel-2 multi-spectral imagery due to its high spatial (10–20 m) and temporal (five days) resolution with support vector machine (SVM) and random forest (RF) classifiers. The species mapped were important for grazing livestock and wildlife (three grass species), indicators of land degradation (one tree genus and one invasive shrub), and a fig tree species. The results show that increasing the number of images, including dry season imagery, results in improved classification accuracy regardless of the classifier (average increase in overall accuracy (OA) = 0.1632). SVM consistently outperformed RF, and the most accurate model and was SVM with a radial kernel using imagery from both wet and dry seasons (OA = 0.8217). Maps showed that seasonal grazing areas provide functionally different grazing opportunities and have different vegetation characteristics that are critical to a landscape’s ability to support large populations of both livestock and wildlife. This study highlights the potential of multi-spectral satellite imagery for species-level mapping of savannas.

A novel classifier for multivariate instance using graph class signatures

Applications like identifying different customers from their unique buying behaviours, determining ratings of a product given by users based on different sets of features, etc. require classification using class-specific subsets of features. Most of the existing state-of-the-art classifiers for multivariate data use complete feature set for classification regardless of the different class labels. Decision tree classifier can produce class-wise subsets of features. However, none of these classifiers model the relationship between features which may enhance classification accuracy. We call the class-specific subsets of features and the features’ interrelationships as class signatures. In this work, we propose to map the original input space of multivariate data to the feature space characterized by connected graphs as graphs can easily model entities, their attributes, and relationships among attributes. Mostly, entities are modeled using graphs, where graphs occur naturally, for example, chemical compounds. However, graphs do not occur naturally in multivariate data. Thus, extracting class signatures from multivariate data is a challenging task. We propose some feature selection heuristics to obtain class-specific prominent subgraph signatures. We also propose two variants of class signatures based classifier namely: 1) maximum matching signature (gMM), and 2) score and size of matched signatures (gSM). The effectiveness of the proposed approach on real-world and synthetic datasets has been studied and compared with other established classifiers. Experimental results confirm the ascendancy of the proposed class signatures based classifier on most of the datasets.

Shape-restricted support vector machine (SR-SVM): a SVM classifier taking supplementary shape information of input

Many classification problems contain shape information from input features, such as monotonic, convex, and concave. In this research, we propose a new classifier, called Shape-Restricted Support Vector Machine (SR-SVM), which takes the component-wise shape information to enhance classification accuracy. There exists vast research literature on monotonic classification covering monotonic or ordinal shapes. Our proposed classifier extends to handle convex and concave types of features, and combinations of these types. While standard SVM uses linear separating hyperplanes, our novel SR-SVM essentially constructs non-parametric and nonlinear separating planes subject to component-wise shape restrictions. We formulate SR-SVM classifier as a convex optimization problem and solve it using an active-set algorithm. The approach applies basis function expansions on the input and effectively utilizes the standard SVM solver. We illustrate our methodology using simulation and real world examples, and show that SR-SVM improves the classification performance with additional shape information of input.

An energy-cyber-physical system for personalized normative messaging interventions: Identification and classification of behavioral reference groups

Within residences, normative messaging interventions have encouraged households to engage in various pro-environmental behaviors. In norm-based intervention campaigns, it is hypothesized that more personally relevant reference groups increase norm adherence, thus improving the effectiveness of normative messaging interventions. Advanced energy grid infrastructure, such as smart meters and cloud computing, enables the creation of highly personalized behavioral reference groups in a non-invasive manner by dynamically classifying households into highly similar user groups based on usage patterns. Unfortunately, it remains unclear how readily available data on household energy use and housing characteristics affect the classification performance of dynamic behavioral reference groups. Therefore, this research evaluates the classification performance of dynamic behavioral reference groups using readily available data. An energy-cyber-physical system for personalized normative messaging interventions is trained and tested using one-year of energy use data from 2248 households in Holland, Michigan. Dynamic behavioral reference group classification proved very accurate, 94.7–95.9% for weekly feedback and 89.9–93.1% for monthly feedback using only readily available data. In addition, using more historical energy use data contributes to enhancing classification accuracy. Lastly, high classification performance for each behavioral reference group is achieved at 97.6% of precision, recall and F1-score. With the proposed system, it is possible to dynamically assign highly personalized behavioral reference groups to households every billing cycle even if behavioral patterns are subject to change. Thus, interveners will be able to deploy personalized normative feedback messages on a large scale.

Multiplanar analysis for pulmonary nodule classification in CT images using deep convolutional neural network and generative adversarial networks.

Early detection and treatment of lung cancer holds great importance. However, pulmonary-nodule classification using CT images alone is difficult to realize. To address this concern, a method for pulmonary-nodule classification based on a deep convolutional neural network (DCNN) and generative adversarial networks (GAN) has previously been proposed by the authors. In that method, the said classification was performed exclusively using axial cross sections of pulmonary nodules. During actual medical-examination procedures, however, a comprehensive judgment can only be made via observation of various pulmonary-nodule cross sections. In the present study, a comprehensive analysis was performed by extending the application of the previously proposed DCNN- and GAN-based automatic classification method to multiple cross sections of pulmonary nodules. Using the proposed method, CT images of 60 cases with confirmed pathological diagnosis by biopsy are analyzed. Firstly, multiplanar images of the pulmonary nodule are generated. Classification training was performed for three DCNNs. A certain pretraining was initially performed using GAN-generated nodule images. This was followed by fine-tuning of each pretrained DCNN using original nodule images provided as input. As a result of the evaluation, the specificity was 77.8% and the sensitivity was 93.9%. Additionally, the specificity was observed to have improved by 11.1% without any reduction in the sensitivity, compared to our previous report. This study reports development of a comprehensive analysis method to classify pulmonary nodules at multiple sections using GAN and DCNN. The effectiveness of the proposed discrimination method based on use of multiplanar images has been demonstrated to be improved compared to that realized in a previous study reported by the authors. In addition, the possibility of enhancing classification accuracy via application of GAN-generated images, instead of data augmentation, for pretraining even for medical datasets that contain relatively few images has also been demonstrated.

ChaosNet: A chaos based artificial neural network architecture for classification.

Inspired by chaotic firing of neurons in the brain, we propose ChaosNet-a novel chaos based artificial neural network architecture for classification tasks. ChaosNet is built using layers of neurons, each of which is a 1D chaotic map known as the Generalized Luröth Series (GLS) that has been shown in earlier works to possess very useful properties for compression, cryptography, and for computing XOR and other logical operations. In this work, we design a novel learning algorithm on ChaosNet that exploits the topological transitivity property of the chaotic GLS neurons. The proposed learning algorithm gives consistently good performance accuracy in a number of classification tasks on well known publicly available datasets with very limited training samples. Even with as low as seven (or fewer) training samples/class (which accounts for less than 0.05% of the total available data), ChaosNet yields performance accuracies in the range of 73.89%-98.33%. We demonstrate the robustness of ChaosNet to additive parameter noise and also provide an example implementation of a two layer ChaosNet for enhancing classification accuracy. We envisage the development of several other novel learning algorithms on ChaosNet in the near future.