Automated Taxonomic Identification of Insects with Expert-Level Accuracy Using Effective Feature Transfer from Convolutional Networks

Abstract Rapid and reliable identification of insects is important in many contexts, from the detection of disease vectors and invasive species to the sorting of material from biodiversity inventories. Because of the shortage of adequate expertise, there has long been an interest in developing automated systems for this task. Previous attempts have been based on laborious and complex handcrafted extraction of image features, but in recent years it has been shown that sophisticated convolutional neural networks (CNNs) can learn to extract relevant features automatically, without human intervention. Unfortunately, reaching expert-level accuracy in CNN identifications requires substantial computational power and huge training data sets, which are often not available for taxonomic tasks. This can be addressed using feature transfer: a CNN that has been pretrained on a generic image classification task is exposed to the taxonomic images of interest, and information about its perception of those images is used in training a simpler, dedicated identification system. Here, we develop an effective method of CNN feature transfer, which achieves expert-level accuracy in taxonomic identification of insects with training sets of 100 images or less per category, depending on the nature of data set. Specifically, we extract rich representations of intermediate to high-level image features from the CNN architecture VGG16 pretrained on the ImageNet data set. This information is submitted to a linear support vector machine classifier, which is trained on the target problem. We tested the performance of our approach on two types of challenging taxonomic tasks: 1) identifying insects to higher groups when they are likely to belong to subgroups that have not been seen previously and 2) identifying visually similar species that are difficult to separate even for experts. For the first task, our approach reached $CDATA[$CDATA[$>$$92% accuracy on one data set (884 face images of 11 families of Diptera, all specimens representing unique species), and $CDATA[$CDATA[$>$$96% accuracy on another (2936 dorsal habitus images of 14 families of Coleoptera, over 90% of specimens belonging to unique species). For the second task, our approach outperformed a leading taxonomic expert on one data set (339 images of three species of the Coleoptera genus Oxythyrea; 97% accuracy), and both humans and traditional automated identification systems on another data set (3845 images of nine species of Plecoptera larvae; 98.6 % accuracy). Reanalyzing several biological image identification tasks studied in the recent literature, we show that our approach is broadly applicable and provides significant improvements over previous methods, whether based on dedicated CNNs, CNN feature transfer, or more traditional techniques. Thus, our method, which is easy to apply, can be highly successful in developing automated taxonomic identification systems even when training data sets are small and computational budgets limited. We conclude by briefly discussing some promising CNN-based research directions in morphological systematics opened up by the success of these techniques in providing accurate diagnostic tools.

Download Full-text

An Incremental Isomap Method for Hyperspectral Dimensionality Reduction and Classification

Photogrammetric Engineering & Remote Sensing ◽

10.14358/pers.87.7.445 ◽

2021 ◽

Vol 87 (6) ◽

pp. 445-455

Author(s):

Yi Ma ◽

Zezhong Zheng ◽

Yutang Ma ◽

Mingcang Zhu ◽

Ran Huang ◽

...

Keyword(s):

Manifold Learning ◽

Nearest Neighbor ◽

Hyperspectral Image ◽

Hyperspectral Data ◽

Training Data ◽

Support Vector ◽

Data Sets ◽

K Nearest Neighbor ◽

Data Set ◽

Data Points

Many manifold learning algorithms conduct an eigen vector analysis on a data-similarity matrix with a size of N×N, where N is the number of data points. Thus, the memory complexity of the analysis is no less than O(N2). We pres- ent in this article an incremental manifold learning approach to handle large hyperspectral data sets for land use identification. In our method, the number of dimensions for the high-dimensional hyperspectral-image data set is obtained with the training data set. A local curvature varia- tion algorithm is utilized to sample a subset of data points as landmarks. Then a manifold skeleton is identified based on the landmarks. Our method is validated on three AVIRIS hyperspectral data sets, outperforming the comparison algorithms with a k–nearest-neighbor classifier and achieving the second best performance with support vector machine.

Download Full-text

A new method for Arabic/Farsi numeral data set size reduction via modified frequency diagram matching

Kybernetes ◽

10.1108/k-10-2013-0226 ◽

2014 ◽

Vol 43 (5) ◽

pp. 817-834

Author(s):

Mohammad Amin Shayegan ◽

Saeed Aghabozorgi

Keyword(s):

Optical Character Recognition ◽

Size Reduction ◽

Training Data ◽

Support Vector ◽

Data Sets ◽

Training Algorithms ◽

Data Set ◽

Content Type ◽

Set Size ◽

Frequency Diagram

Purpose – Pattern recognition systems often have to handle problem of large volume of training data sets including duplicate and similar training samples. This problem leads to large memory requirement for saving and processing data, and the time complexity for training algorithms. The purpose of the paper is to reduce the volume of training part of a data set – in order to increase the system speed, without any significant decrease in system accuracy. Design/methodology/approach – A new technique for data set size reduction – using a version of modified frequency diagram approach – is presented. In order to reduce processing time, the proposed method compares the samples of a class to other samples in the same class, instead of comparing samples from different classes. It only removes patterns that are similar to the generated class template in each class. To achieve this aim, no feature extraction operation was carried out, in order to produce more precise assessment on the proposed data size reduction technique. Findings – The results from the experiments, and according to one of the biggest handwritten numeral standard optical character recognition (OCR) data sets, Hoda, show a 14.88 percent decrease in data set volume without significant decrease in performance. Practical implications – The proposed technique is effective for size reduction for all pictorial databases such as OCR data sets. Originality/value – State-of-the-art algorithms currently used for data set size reduction usually remove samples near to class's centers, or support vector (SV) samples between different classes. However, the samples near to a class center have valuable information about class characteristics, and they are necessary to build a system model. Also, SV s are important samples to evaluate the system efficiency. The proposed technique, unlike the other available methods, keeps both outlier samples, as well as the samples close to the class centers.

Download Full-text

Comparison of Support Vector Machine and Random Forest Algorithms for Invasive and Expansive Species Classification Using Airborne Hyperspectral Data

Remote Sensing ◽

10.3390/rs12030516 ◽

2020 ◽

Vol 12 (3) ◽

pp. 516 ◽

Cited By ~ 7

Author(s):

Anita Sabat-Tomala ◽

Edwin Raczko ◽

Bogdan Zagajewski

Keyword(s):

Support Vector Machine ◽

Random Forest ◽

Hyperspectral Data ◽

Machine Learning Algorithms ◽

Training Data ◽

Support Vector ◽

Data Sets ◽

Natural Ecosystem ◽

Data Set ◽

Calamagrostis Epigejos

Invasive and expansive plant species are considered a threat to natural biodiversity because of their high adaptability and low habitat requirements. Species investigated in this research, including Solidago spp., Calamagrostis epigejos, and Rubus spp., are successfully displacing native vegetation and claiming new areas, which in turn severely decreases natural ecosystem richness, as they rapidly encroach on protected areas (e.g., Natura 2000 habitats). Because of the damage caused, the European Union (EU) has committed all its member countries to monitor biodiversity. In this paper we compared two machine learning algorithms, Support Vector Machine (SVM) and Random Forest (RF), to identify Solidago spp., Calamagrostis epigejos, and Rubus spp. on HySpex hyperspectral aerial images. SVM and RF are reliable and well-known classifiers that achieve satisfactory results in the literature. Data sets containing 30, 50, 100, 200, and 300 pixels per class in the training data set were used to train SVM and RF classifiers. The classifications were performed on 430-spectral bands and on the most informative 30 bands extracted using the Minimum Noise Fraction (MNF) transformation. As a result, maps of the spatial distribution of analyzed species were achieved; high accuracies were observed for all data sets and classifiers (an average F1 score above 0.78). The highest accuracies were obtained using 30 MNF bands and 300 sample pixels per class in the training data set (average F1 score > 0.9). Lower training data set sample sizes resulted in decreased average F1 scores, up to 13 percentage points in the case of 30-pixel samples per class.

Download Full-text

Reconstruct the Support Vectors to Improve LSSVM Sparseness for Mill Load Prediction

Mathematical Problems in Engineering ◽

10.1155/2017/4191789 ◽

2017 ◽

Vol 2017 ◽

pp. 1-12

Author(s):

Gangquan Si ◽

Jianquan Shi ◽

Zhang Guo ◽

Lixin Jia ◽

Yanbin Zhang

Keyword(s):

Synthetic Data ◽

Training Model ◽

Least Square ◽

Training Data ◽

Support Vector ◽

Data Sets ◽

Load Prediction ◽

Data Set ◽

Support Vectors ◽

Reconstructed Data

The sparse strategy plays a significant role in the application of the least square support vector machine (LSSVM), to alleviate the condition that the solution of LSSVM is lacking sparseness and robustness. In this paper, a sparse method using reconstructed support vectors is proposed, which has also been successfully applied to mill load prediction. Different from other sparse algorithms, it no longer selects the support vectors from training data set according to the ranked contributions for optimization of LSSVM. Instead, the reconstructed data is obtained first based on the initial model with all training data. Then, select support vectors from reconstructed data set according to the location information of density clustering in training data set, and the process of selecting is terminated after traversing the total training data set. Finally, the training model could be built based on the optimal reconstructed support vectors and the hyperparameter tuned subsequently. What is more, the paper puts forward a supplemental algorithm to subtract the redundancy support vectors of previous model. Lots of experiments on synthetic data sets, benchmark data sets, and mill load data sets are carried out, and the results illustrate the effectiveness of the proposed sparse method for LSSVM.

Download Full-text

RF-Identity

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies ◽

10.1145/3448101 ◽

2021 ◽

Vol 5 (1) ◽

pp. 1-23

Author(s):

Chao Feng ◽

Jie Xiong ◽

Liqiong Chang ◽

Fuwei Wang ◽

Ju Wang ◽

...

Keyword(s):

Data Augmentation ◽

Critical Role ◽

Training Data ◽

Identification Accuracy ◽

Identification System ◽

Frequency Bandwidth ◽

Person Identification ◽

Data Set ◽

Rf Devices ◽

Large Frequency

Person identification plays a critical role in a large range of applications. Recently, RF based person identification becomes a hot research topic due to the contact-free nature of RF sensing that is particularly appealing in current COVID-19 pandemic. However, existing systems still have multiple limitations: i) heavily rely on the gait patterns of users for identification; ii) require a large amount of data to train the model and also extensive retraining for new users and iii) require a large frequency bandwidth which is not available on most commodity RF devices for static person identification. This paper proposes RF-Identity, an RFID-based identification system to address the above limitations and the contribution is threefold. First, by integrating walking pattern features with unique body shape features (e.g., height), RF-Identity achieves a high accuracy in person identification. Second, RF-Identity develops a data augmentation scheme to expand the size of the training data set, thus reducing the human effort in data collection. Third, RF-Identity utilizes the tag diversity in spatial domain to identify static users without a need of large frequency bandwidth. Extensive experiments show an identification accuracy of 94.2% and 95.9% for 50 dynamic and static users, respectively.

Download Full-text

A sentiment analysis system for social media using machine learning techniques: Social enablement

Digital Scholarship in the Humanities ◽

10.1093/llc/fqy037 ◽

2018 ◽

Vol 34 (3) ◽

pp. 569-581 ◽

Cited By ~ 1

Author(s):

Sujata Rani ◽

Parteek Kumar

Keyword(s):

Machine Learning ◽

Social Media ◽

Sentiment Analysis ◽

Media Analysis ◽

Training Data ◽

Machine Learning Techniques ◽

Support Vector ◽

Analysis Tool ◽

Data Set ◽

Learning Techniques

Abstract In this article, an innovative approach to perform the sentiment analysis (SA) has been presented. The proposed system handles the issues of Romanized or abbreviated text and spelling variations in the text to perform the sentiment analysis. The training data set of 3,000 movie reviews and tweets has been manually labeled by native speakers of Hindi in three classes, i.e. positive, negative, and neutral. The system uses WEKA (Waikato Environment for Knowledge Analysis) tool to convert these string data into numerical matrices and applies three machine learning techniques, i.e. Naive Bayes (NB), J48, and support vector machine (SVM). The proposed system has been tested on 100 movie reviews and tweets, and it has been observed that SVM has performed best in comparison to other classifiers, and it has an accuracy of 68% for movie reviews and 82% in case of tweets. The results of the proposed system are very promising and can be used in emerging applications like SA of product reviews and social media analysis. Additionally, the proposed system can be used in other cultural/social benefits like predicting/fighting human riots.

Download Full-text

Data Analysis With Shapley Values For Automatic Subject Selection in Alzheimer's Disease Data Sets Using Interpretable Machine Learning

10.21203/rs.3.rs-245707/v1 ◽

2021 ◽

Author(s):

Louise Bloch ◽

Christoph M. Friedrich

Keyword(s):

Alzheimer’S Disease ◽

Alzheimer's Disease ◽

Test Data ◽

Noisy Data ◽

Training Data ◽

Data Sets ◽

Data Set ◽

Model Interpretation ◽

Percentage Points ◽

Shapley Values

Abstract Background: The prediction of whether Mild Cognitive Impaired (MCI) subjects will prospectively develop Alzheimer's Disease (AD) is important for the recruitment and monitoring of subjects for therapy studies. Machine Learning (ML) is suitable to improve early AD prediction. The etiology of AD is heterogeneous, which leads to noisy data sets. Additional noise is introduced by multicentric study designs and varying acquisition protocols. This article examines whether an automatic and fair data valuation method based on Shapley values can identify subjects with noisy data. Methods: An ML-workow was developed and trained for a subset of the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort. The validation was executed for an independent ADNI test data set and for the Australian Imaging, Biomarker and Lifestyle Flagship Study of Ageing (AIBL) cohort. The workow included volumetric Magnetic Resonance Imaging (MRI) feature extraction, subject sample selection using data Shapley, Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) for model training and Kernel SHapley Additive exPlanations (SHAP) values for model interpretation. This model interpretation enables clinically relevant explanation of individual predictions. Results: The XGBoost models which excluded 116 of the 467 subjects from the training data set based on their Logistic Regression (LR) data Shapley values outperformed the models which were trained on the entire training data set and which reached a mean classification accuracy of 58.54 % by 14.13 % (8.27 percentage points) on the independent ADNI test data set. The XGBoost models, which were trained on the entire training data set reached a mean accuracy of 60.35 % for the AIBL data set. An improvement of 24.86 % (15.00 percentage points) could be reached for the XGBoost models if those 72 subjects with the smallest RF data Shapley values were excluded from the training data set. Conclusion: The data Shapley method was able to improve the classification accuracies for the test data sets. Noisy data was associated with the number of ApoEϵ4 alleles and volumetric MRI measurements. Kernel SHAP showed that the black-box models learned biologically plausible associations.

Download Full-text

Simple Convolutional-Based Models: Are They Learning the Task or the Data?

Neural Computation ◽

10.1162/neco_a_01446 ◽

2021 ◽

pp. 1-17

Author(s):

Luis Sa-Couto ◽

Andreas Wichert

Keyword(s):

Neural Networks ◽

Pattern Recognition ◽

Training Data ◽

Model Complexity ◽

Data Sets ◽

Simple Task ◽

Data Set ◽

Knowing That ◽

Handwritten Digit ◽

End To End

Abstract Convolutional neural networks (CNNs) evolved from Fukushima's neocognitron model, which is based on the ideas of Hubel and Wiesel about the early stages of the visual cortex. Unlike other branches of neocognitron-based models, the typical CNN is based on end-to-end supervised learning by backpropagation and removes the focus from built-in invariance mechanisms, using pooling not as a way to tolerate small shifts but as a regularization tool that decreases model complexity. These properties of end-to-end supervision and flexibility of structure allow the typical CNN to become highly tuned to the training data, leading to extremely high accuracies on typical visual pattern recognition data sets. However, in this work, we hypothesize that there is a flip side to this capability, a hidden overfitting. More concretely, a supervised, backpropagation based CNN will outperform a neocognitron/map transformation cascade (MTCCXC) when trained and tested inside the same data set. Yet if we take both models trained and test them on the same task but on another data set (without retraining), the overfitting appears. Other neocognitron descendants like the What-Where model go in a different direction. In these models, learning remains unsupervised, but more structure is added to capture invariance to typical changes. Knowing that, we further hypothesize that if we repeat the same experiments with this model, the lack of supervision may make it worse than the typical CNN inside the same data set, but the added structure will make it generalize even better to another one. To put our hypothesis to the test, we choose the simple task of handwritten digit classification and take two well-known data sets of it: MNIST and ETL-1. To try to make the two data sets as similar as possible, we experiment with several types of preprocessing. However, regardless of the type in question, the results align exactly with expectation.

Download Full-text

Exploiting Rules to Enhance Machine Learning in Extracting Information From Multi-Institutional Prostate Pathology Reports

JCO Clinical Cancer Informatics ◽

10.1200/cci.20.00028 ◽

2020 ◽

pp. 865-874

Author(s):

Enrico Santus ◽

Tal Schuster ◽

Amir M. Tahmasebi ◽

Clara Li ◽

Adam Yala ◽

...

Keyword(s):

Machine Learning ◽

Hybrid Systems ◽

High Performance ◽

Feature Model ◽

Training Data ◽

Gradient Boosting ◽

Support Vector ◽

Data Set ◽

Extreme Gradient Boosting ◽

Pathology Reports

PURPOSE Literature on clinical note mining has highlighted the superiority of machine learning (ML) over hand-crafted rules. Nevertheless, most studies assume the availability of large training sets, which is rarely the case. For this reason, in the clinical setting, rules are still common. We suggest 2 methods to leverage the knowledge encoded in pre-existing rules to inform ML decisions and obtain high performance, even with scarce annotations. METHODS We collected 501 prostate pathology reports from 6 American hospitals. Reports were split into 2,711 core segments, annotated with 20 attributes describing the histology, grade, extension, and location of tumors. The data set was split by institutions to generate a cross-institutional evaluation setting. We assessed 4 systems, namely a rule-based approach, an ML model, and 2 hybrid systems integrating the previous methods: a Rule as Feature model and a Classifier Confidence model. Several ML algorithms were tested, including logistic regression (LR), support vector machine (SVM), and eXtreme gradient boosting (XGB). RESULTS When training on data from a single institution, LR lags behind the rules by 3.5% (F1 score: 92.2% v 95.7%). Hybrid models, instead, obtain competitive results, with Classifier Confidence outperforming the rules by +0.5% (96.2%). When a larger amount of data from multiple institutions is used, LR improves by +1.5% over the rules (97.2%), whereas hybrid systems obtain +2.2% for Rule as Feature (97.7%) and +2.6% for Classifier Confidence (98.3%). Replacing LR with SVM or XGB yielded similar performance gains. CONCLUSION We developed methods to use pre-existing handcrafted rules to inform ML algorithms. These hybrid systems obtain better performance than either rules or ML models alone, even when training data are limited.

Download Full-text

Artificial bee colony algorithm for feature selection and improved support vector machine for text classification

Information Discovery and Delivery ◽

10.1108/idd-09-2018-0045 ◽

2019 ◽

Vol 47 (3) ◽

pp. 154-170

Author(s):

Janani Balakumar ◽

S. Vijayarani Mohan

Keyword(s):

Support Vector Machine ◽

Feature Selection ◽

Text Classification ◽

Support Vector ◽

Data Sets ◽

Selection Algorithm ◽

Data Set ◽

Content Type ◽

Benchmark Data ◽

Bee Colony

Purpose Owing to the huge volume of documents available on the internet, text classification becomes a necessary task to handle these documents. To achieve optimal text classification results, feature selection, an important stage, is used to curtail the dimensionality of text documents by choosing suitable features. The main purpose of this research work is to classify the personal computer documents based on their content. Design/methodology/approach This paper proposes a new algorithm for feature selection based on artificial bee colony (ABCFS) to enhance the text classification accuracy. The proposed algorithm (ABCFS) is scrutinized with the real and benchmark data sets, which is contrary to the other existing feature selection approaches such as information gain and χ2 statistic. To justify the efficiency of the proposed algorithm, the support vector machine (SVM) and improved SVM classifier are used in this paper. Findings The experiment was conducted on real and benchmark data sets. The real data set was collected in the form of documents that were stored in the personal computer, and the benchmark data set was collected from Reuters and 20 Newsgroups corpus. The results prove the performance of the proposed feature selection algorithm by enhancing the text document classification accuracy. Originality/value This paper proposes a new ABCFS algorithm for feature selection, evaluates the efficiency of the ABCFS algorithm and improves the support vector machine. In this paper, the ABCFS algorithm is used to select the features from text (unstructured) documents. Although, there is no text feature selection algorithm in the existing work, the ABCFS algorithm is used to select the data (structured) features. The proposed algorithm will classify the documents automatically based on their content.

Download Full-text