Probabilistic Artificial Neural Network Research Articles

Postprocessing of NWP Precipitation Forecasts Using Deep Learning

Abstract This study analyzes the potential of deep learning using probabilistic artificial neural networks (ANNs) for postprocessing ensemble precipitation forecasts at four observation locations. We split the precipitation forecast problem into two tasks: estimating the probability of precipitation and predicting the hourly precipitation. We then compare the performance with classical statistical postprocessing (logistical regression and GLM). ANNs show a higher performance at three of the four stations for estimating the probability of precipitation and at all stations for predicting the hourly precipitation. Further, two more general ANN models are trained using the merged data from all four stations. These general ANNs exhibit an increase in performance compared to the station-specific ANNs at most stations. However, they show a significant decay in performance at one of the stations at estimating the hourly precipitation. The general models seem capable of learning meaningful interactions in the data and generalizing these to improve the performance at other sites, which also causes the loss of local information at one station. Thus, this study indicates the potential of deep learning in weather forecasting workflows.

Mid-infrared spectroscopy combined with multivariate analysis and machine-learning: A powerful tool to simultaneously assess geographical origin, growing conditions and bitter content in Gentiana lutea roots

Mid-infrared spectroscopy was explored in order to evaluate its ability to classify geographical origin and bitter content of Gentiana lutea roots sourced from wild and cultivated growing conditions. Wild and cultivated Gentiana lutea roots from the four French mountains Massif Central, Jura, Alpes and Pyrénées were analyzed by Infrared spectroscopy and liquid chromatography. Unsupervised analyses assessed heterogeneity of Gentiana lutea roots in the different sampling sites due to their evolutive composition along plant growth. Predictive models using partial least squares discriminant analysis and probabilistic artificial neural network were discussed according to FTIR spectral regions and gave 100 % accuracy in authentication of geographical origin and growing conditions. Nevertheless, classification of gentian roots according to their bitter content, based on FTIR spectral signatures, was challenging due to their biological heterogeneity. We propose an unprecedented classification of Gentiana lutea roots according to their analyzed bitter content: Low, Medium and High, comprised in the range [6–8] %, [8–10] % and [10–12] % in dry weight, respectively. FTIR coupled with chemometrics applied directly on gentian roots allowed for a decent level of predictability by PLS-DA (Q2Cum = 0.41) and Artificial Neural Network (89.1 % accuracy), when using the (650–1800 cm−1) infrared region.

Probabilistic Artificial Neural Network for Line-Edge-Roughness-Induced Random Variation in FinFET

Line-edge-roughness (LER) is one of undesirable process-induced random variation sources. LER is mostly occurred in the process of photo-lithography and etching, and it provokes random variation in performance of transistors such as metal oxide semiconductor field effect transistor (MOSFET), fin-shaped field effect transistor (FinFET), and gate-all-around field effect transistor (GAAFET). LER was analyzed/characterized with technology computer-aided design (TCAD), but it is fundamentally very time consuming. To tackle this issue, machine learning (ML)-based method is proposed in this work. LER parameters (i.e., amplitude, and correlation length X, Y) are provided as inputs. Then, artificial neural network (ANN) predicts 7-parameters [i.e., off-state leakage current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">off</sub> ), saturation drain current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dsat</sub> ), linear drain current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dlin</sub> ), low drain current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dlo</sub> ), high drain current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dhi</sub> ), saturation threshold voltage (V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tsat</sub> ), and linear threshold voltage (V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tlin</sub> )] which are usually used to evaluate the performance of FinFET. First, how datasets for training process of ANN were generated is explained. Next, the evaluation method for probabilistic problem is introduced. Finally, the architecture of ANN, training process and our new proposition is presented. It turned out that the prediction results (i.e., non-Gaussian distribution of device performance metrics) obtained from the ANN were very similar to that from TCAD in the respect of both qualitative and quantitative comparison.

Cancer Risk Analysis Based on Improved Probabilistic Neural Network.

The problem of cancer risk analysis is of great importance to health-service providers and medical researchers. In this study, we propose a novel Artificial Neural Network (ANN) algorithm based on the probabilistic framework, which aims to investigate patient patterns associated with their disease development. Compared to the traditional ANN where input features are directly extracted from raw data, the proposed probabilistic ANN manipulates original inputs according to their probability distribution. More precisely, the Naïve Bayes and Markov chain models are used to approximate the posterior distribution of the raw inputs, which provides a useful estimation of subsequent disease development. Later, this distribution information is further leveraged as additional input to train ANN. Additionally, to reduce the training cost and to boost the generalization capability, a sparse training strategy is also introduced. Experimentally, one of the largest cancer-related datasets is employed in this study. Compared to state-of-the-art methods, the proposed algorithm achieves a much better outcome, in terms of the prediction accuracy of subsequent disease development. The result also reveals the potential impact of patients' disease sequence on their future risk management.

Using Coherence-based spectro-spatial filters for stimulus features prediction from electro-corticographic recordings

The traditional approach in neuroscience relies on encoding models where brain responses are related to different stimuli in order to establish dependencies. In decoding tasks, on the contrary, brain responses are used to predict the stimuli, and traditionally, the signals are assumed stationary within trials, which is rarely the case for natural stimuli. We hypothesize that a decoding model assuming each experimental trial as a realization of a random process more likely reflects the statistical properties of the undergoing process compared to the assumption of stationarity. Here, we propose a Coherence-based spectro-spatial filter that allows for reconstructing stimulus features from brain signal’s features. The proposed method extracts common patterns between features of the brain signals and the stimuli that produced them. These patterns, originating from different recording electrodes are combined, forming a spatial filter that produces a unified prediction of the presented stimulus. This approach takes into account frequency, phase, and spatial distribution of brain features, hence avoiding the need to predefine specific frequency bands of interest or phase relationships between stimulus and brain responses manually. Furthermore, the model does not require the tuning of hyper-parameters, reducing significantly the computational load attached to it. Using three different cognitive tasks (motor movements, speech perception, and speech production), we show that the proposed method consistently improves stimulus feature predictions in terms of correlation (group averages of 0.74 for motor movements, 0.84 for speech perception, and 0.74 for speech production) in comparison with other methods based on regularized multivariate regression, probabilistic graphical models and artificial neural networks. Furthermore, the model parameters revealed those anatomical regions and spectral components that were discriminant in the different cognitive tasks. This novel method does not only provide a useful tool to address fundamental neuroscience questions, but could also be applied to neuroprosthetics.

Probabilistic artificial neural network and E-nose based classification of Rhyzopertha dominica infestation in stored rice grains

The versatility of artificial neural network with back propagation multilayer perceptron approach could entitle an easy and methodical interpretation of results corresponding to multiple metal oxides sensor in an electronic nose. Three algorithms discriminant factorial analysis (DFA), soft independent modeling by class analogy (SIMCA), probabilistic artificial neural network (PANN) with back propagation multilayer perceptron (BPNN) were used for the classification of R. dominica infested rice stored for 225 days via 18 metal oxide sensors in E-nose. The coefficient of correlation for the three approaches were 88 (DFA), 96 (SIMCA), 98.96 (BPNN) %, respectively. The percentage discrimination index was more distinct between 0 and 225 days R. dominica infested rice (98%) than 0–180 days (93%), and 0–135 days (88%). The residual errors of validation and cross validation for SIMCA were 1.04 × 10−3 and 1.26 × 10−3 respectively. Major metal oxide sensors responsible for the production of volatiles were P30/1, T 30/1, PA/2, P30/2, T70/2, P40/1, and P40/2. The overall relative errors during artificial neural network training and testing were 0.092 and 0.286 respectively. The artificial neural network relative error for scale dependents in response to metal oxide sensors for mean, SD, % RSD were 0.033, 0.162, 0.081, respectively. The applicability of E-nose with neural network could help in securing the data analysis time without loss of information and can also work well for noxious odors which might not be able to be categorized by human olfactory.

An Efficient Image Segmentation and Classification of Tumors in Mammography Image using K-Means Clustering with ROI and ANN Classifier

Breast cancer is commonly occurred in women that causes of death. An early detection of breast cancer can be used to long survival of patients. The tumor is detected in breast using mammography X-ray image. This mammography is one of the special cases of CT scan that is utilized to discover the breast cancer. It is screening method to find the malignant tumor cells in breast among women at early stage to avoid the deaths of patient. However, the radiologists not forever provide accurate results and detection of malignant tumor in breast is a one of the challenging issue due to the tumor cells structure. Therefore, this paper presents K-means clustering with Probabilistic neural networks for improvement of mammogram image quality to provide accurate results. K- Means clustering is applied to segment the mammogram image efficiently and presents an automatic support system for image classification using trained probabilistic neural networks and Artificial Neural networks (ANN) classification will be employed to classify the stage of breast cancer tumor that is caring, malignant or standard. Describing the ROI (Region of Interest), utilizing Region growing segmentation technique, evaluating the area of assumed cancer and establish the image classification of the area on the mammogram image using Sample K-Means Clustering and PNN techniques. Gray Level Co-occurrence Matrix (GLCM) is to extract the features from the texture that features may be in the form of structure, location and outside etc. In this process, the preprocessing method is applied to reduce the noise from the mammogram image for the image enhancement. For the preprocessing, the filter methods are used to give image quality. The performance of k-means clustering and PNN techniques provides fast and accurate image segmentation and classification than the other techniques to detect tumor cells in breast at early stage efficiently.

The use of a non-probabilistic artificial neural network to consider uncertainties in vibration-based-damage detection

The effectiveness of artificial neural networks (ANNs) when applied to pattern recognition in vibration-based damage detection has been demonstrated in many studies because they are capable of providing accurate results and the reliable identification of structural damage based on modal data. However, the use of ANNs has been questioned in terms of its reliability in the face of uncertainties in measurement and modeling data. Attempts to incorporate a probabilistic method into an ANN by treating the uncertainties as normally distributed random variables has delivered promising solutions to this problem, but the probabilistic method is less straightforward in practice because it is often not possible to obtain unbiased probabilistic distributions of the uncertainties. Moreover, the probabilistic ANN method is computationally complex, especially when generating output data. In this study, a non-probabilistic ANN is proposed to address the problem of uncertainty in vibration damage detection using ANNs. The input data for the network consist of natural frequencies and mode shapes, and the output is the Young's modulus (E values), which acts as an elemental stiffness parameter (ESP). Through the interval analysis method, the noise in measured frequencies and mode shapes are considered to be coupled rather than statistically distributed. This method calculates the interval bound (lower and upper bounds) of the ESP changes based on an interval analysis method. The ANN is used to predict the output of this interval bound by considering the uncertainties in the input parameters. To establish the relationship between the input parameters and output parameters, a possibility of damage existence (PoDE) parameter is defined for the undamaged and damaged states. A stiffness reduction factor (SRF) is also used to represent changes in the stiffness parameter. A numerical model and a laboratory-tested steel portal frame demonstrate the efficacy of the method in improving the accuracy of the ANN in the presence of uncertainties. The effect of different severity levels and the influence of different noise levels on the identification results are discussed.

Developing a methodology for early leakage detection in landfills: application of the fuzzy transformation technique and probabilistic artificial neural networks

In this paper, a new methodology is developed for early detecting leakage from landfills to groundwater systems. The optimal number and locations of monitoring wells are determined with respect to maximizing the probability of detecting a pollution source and minimizing the total cost of monitoring system. The concentration of a water quality indicator is simulated by analytically solving the mass transport equations in porous media. In the developed methodology, the uncertainties associated with the leakage characteristics as well as parameters related to pollution mass transport equations are addressed using the fuzzy set theory. Also, a fuzzy transformation technique is utilized, which decomposes the fuzzy membership functions into a number of intervals called $$ \alpha $$ -cuts. Corresponding to each leakage scenario, fuzzy membership function of the concentration of the water quality indicator at each monitoring well is obtained using the fuzzy transformation technique. The resulted fuzzy numbers are compared with a fuzzy threshold for detecting pollution using a fuzzy ranking technique. Finally, the optimum number of monitoring wells with their best locations can be determined using a trade-off curve between the probability of pollution detection and the number of monitoring wells. For real-time operation of the monitoring system, a probabilistic artificial neural network is trained and verified which provides the probability of occurring leakage at each section of the landfill based on the concentration of the water quality indicator at the monitoring wells. The applicability and effectiveness of the proposed methodology in detecting pollution leakage from landfills are examined by applying it to a case study in Iran.

Intelligent Classifier Approach for Prediction and Sensitivity Analysis of Differential Pipe Sticking: A Comparative Study

Prediction of differential pipe sticking (DPS) prior to occurrence, and taking preventive measures, is one of the best approaches to minimize the risk of DPS. In this paper, probabilistic artificial neural network (ANN) has been introduced. Moreover, conventional ANNs through multilayer perceptron (MLP) and radial basis function (RBF) have been used to compare with probabilistic ANN. Furthermore, to determine the most important parameters, forward selection sensitivity analysis has been applied. By predicting DPS and performing sensitivity analysis, it is possible to improve well planning process. The results from the analyses have shown the better potentiality of the probabilistic ANN in this area.

A New Artificial Neural Networks Approach for Diagnosing Diabetes Disease Type II

Diabetes is one of the major health problems as it causes physical disability and even death in people. Therefore, to diagnose this dangerous disease better, methods with minimum error rate must be used. Different models of artificial neural networks have the capability to diagnose this disease with minimum error. Hence, in this paper we have used probabilistic artificial neural networks for an approach to diagnose diabetes disease type II. We took advantage of Pima Indians Diabetes dataset with 768 samples in our experiments. According to this dataset, PNN is implemented in MATLAB. Furthermore, maximizing accuracy of diagnosing the Diabetes disease type II in training and testing the Pima Indians Diabetes dataset is the performance measure in this paper. Finally, we concluded that training accuracy and testing accuracy of the proposed method is 89.56% and 81.49%, respectively

A Novel Probabilistic Artificial Neural Networks Approach for Diagnosing Heart Disease

A Novel Probabilistic Artificial Neural Networks Approach for Diagnosing Heart Disease

A novel formed classifier for analyzing cellular contents in cells images

Several approaches for cell image segmentation have been proposed and applied with success in the biomedical sciences field, but cell content estimation is still a not solved problem. This paper presents a novel formed classifier that combine appropriately “intelligent scissors” (livewire), object’s pixel information (RGB values of colors) and probabilistic artificial neural networks (ANN) techniques for segmenting and estimating precisely the cytoplasm composition in relation to protein, lipid, and carbohydrates. Based on this it is possible to estimate the gonad’s state of development and the maturation stage of oocytes in adult ovigerous females gonads.The ovigerous mature females (identifiable by their brood pouch – empty, with eggs, with eyeless or eyed embryos) were selected under a dissecting microscope in the laboratory. A total of 170 gonadic masses from females containing oocytes in different developmental stages were analyzed. A total of 100 sections were observed by light microscopy in order to record the arrangement of the ovaries and determine the maturity stage of the various germ cells.With the ANN network (Fig. 1) trained is possible to carry out the classification step. Classification aims to associate each cytoplasm pixel to one of the selected classes. In this process each cytoplasm pixel is simulated (Figure 2 A). As result a new image containing the classified cytoplasm is produced, where each class is represented by a different colour level (Figure 2 B). Figure 2 C show from left to right the full “classification flow process” including ROI selection, cytoplasm and nucleus contour extraction, cytoplasm segmentation and cytoplasm content classification. The proposed classifier was tested successfully in a representative image dataset, demonstrating the effectiveness and robustness of this new method facilitating the histochemical analysis of cell contents, based on a gonads model.

Antioxidant activity prediction and classification of some teas using artificial neural networks

In order to characterise and to classify some teas a simple, rapid and economical method based on composition, antioxidant activity and artificial neural networks (ANNs) is proposed. For these purpose two types of ANN based applications have been developed: one for predicting the antioxidant activity and a second one for establishing the class of the teas. The complex relationship between the total antioxidant activity (AA) depending on the total flavonoids content (F), total catechins content (C) and total methyl-xanthines content (MX) of commercial teas was revealed by the first designed feed-forward ANN. Secondly, using a probabilistic ANN, successful tea classification in various classes (green tea, black tea and express black tea) was also performed.

Artificial neural networks for predicting diffusible hydrogen content and cracking susceptibility in rutile flux-cored arc welds

The aim of the current work was to build and validate a back propagation artificial neural network (BPANN) to predict the level of diffusible hydrogen ( H D) based on a variety of flux-cored arc welding (FCAW) parameters such as welding current, contact-tip to work piece distance (CTWD) and shielding gas type for single run, horizontal bead-on-plate steel welds using seamed and seamless rutile consumable wires (Model 1). A probabilistic artificial neural network model (PANN) was then applied to determine the likelihood of hydrogen assisted cold cracking (HACC) in a gapped bead-on-plate (GBOP) test based on the level of H D and welding preheat temperature (Model 2). This approach could be applied to ‘real life’ situations with the availability of sufficient data. Results are also presented for a sensitivity analysis for Model 1, which identifies the key input variables and the effect they have on the output, H D. It was determined that the consumable type and the CTWD were the key variables. A seamless consumable with a H5 classification and an increase in the CTWD, as well as increases in current led mostly to a decrease in the level of diffusible hydrogen. Additionally, Model 1 predicted that using a 25%CO 2–75%Ar shielding gas, instead of 100%CO 2, and an increase in absolute humidity (a function of relative humidity and temperature) in the surrounding atmosphere led to an increase in diffusible hydrogen deposited into the weld metal (WM). Finally, based on the predictions of Model 2 it was established that for the given welding conditions hydrogen cracking in a GBOP test could be avoided by maintaining hydrogen levels at less than ∼3 mL/100 g or using a preheat temperature slightly higher than 100 °C. Analysis of the data also led to the conclusion that further data were required for the GBOP tests when the hydrogen levels were between 4 and 8 mL/100 g to correct and improve confidence in this part of the model.

Probabilistic Artificial Neural Network For Recognizing the Arabic Hand Written Characters

The objective of this study was to present a new technique assists in developing a recognition system for handling the Arabic Hand Written text. The proposed system is called Arabic Hand Written Optical Character Recognition (AHOCR). AHOCR was concerned with recognition of hand written Alphanumeric Arabic characters. In the present work, extracted characters are represented by using geometric moment invariant of order three. The advantage of using moment invariant for pattern classification as compared to the other methods is its invariant with respect to its: position , size and rotation .The proposed technique was divided into three major steps : the first step was concerned with digitization and preprocessing documents to create connect components, detect the skew of characters and correct it .The second step deals with how to use geometric moment invariant features of the input Arabic characters to extract features . The third step focused on description of an advanced system of classification using Probabilistic Neural networks structure which yields significant speed improvement. Our final results indicate and clarify that the proposed AHOCR technique achieves an excellent test accuracy of recognition rated up to 97% for isolated Arabic characters and 96% for Arabic text.

Identification of A/H5N1 Influenza Viruses Using a Single Gene Diagnostic Microarray

In previous work, a simple diagnostic DNA microarray that targeted only the matrix gene segment of influenza A (MChip) was developed and evaluated with patient samples. In this work, the analytical utility of the MChip for detection and subtyping of an emerging virus was evaluated with a diverse set of A/H5N1 influenza viruses. A total of 43 different highly pathogenic A/H5N1 viral isolates that were collected from diverse geographic locations, including Vietnam, Nigeria, Indonesia, and Kazakhstan, representing human, feline, and a variety of avian infections spanning the time period 2003-2006 were used in this study. A probabilistic artificial neural network was developed for automated microarray image interpretation through pattern recognition. The microarray assay and subsequent subtype assignment by the artificial neural network resulted in correct identification of 24 "unknown" A/H5N1 positive samples with no false positives. Analysis of a data set composed of A/H5N1, A/H3N2, and A/H1N1 positive samples and negative controls resulted in a clinical sensitivity of 97% and a clinical specificity of 100%.

Amino Acid Distribution in Meteorites: Diagenesis, Extraction Methods, and Standard Metrics in the Search for Extraterrestrial Biosignatures

The relative abundance of the protein amino acids has been previously investigated as a potential marker for biogenicity in meteoritic samples. However, these investigations were executed without a quantitative metric to evaluate distribution variations, and they did not account for the possibility of interdisciplinary systematic error arising from inter-laboratory differences in extraction and detection techniques. Principal component analysis (PCA), hierarchical cluster analysis (HCA), and stochastic probabilistic artificial neural networks (ANNs) were used to compare the distributions for nine protein amino acids previously reported for the Murchison carbonaceous chondrite, Mars meteorites (ALH84001, Nakhla, and EETA79001), prebiotic synthesis experiments, and terrestrial biota and sediments. These techniques allowed us (1) to identify a shift in terrestrial amino acid distributions secondary to diagenesis; (2) to detect differences in terrestrial distributions that may be systematic differences between extraction and analysis techniques in biological and geological laboratories; and (3) to determine that distributions in meteoritic samples appear more similar to prebiotic chemistry samples than they do to the terrestrial unaltered or diagenetic samples. Both diagenesis and putative interdisciplinary differences in analysis complicate interpretation of meteoritic amino acid distributions. We propose that the analysis of future samples from such diverse sources as meteoritic influx, sample return missions, and in situ exploration of Mars would be less ambiguous with adoption of standardized assay techniques, systematic inclusion of assay standards, and the use of a quantitative, probabilistic metric. We present here one such metric determined by sequential feature extraction and normalization (PCA), information-driven automated exploration of classification possibilities (HCA), and prediction of classification accuracy (ANNs).

Evaluation of chemometric techniques and artificial neural networks for cancer screening using Cu, Fe, Se and Zn concentrations in blood serum

It is known that variations in the concentrations of certain trace elements in bodily fluids may be an indication of an alteration of the organism's normal functioning. Multivariate analysis of such data, and its comparison against proper reference values, can thus be employed as possible screening tests. Such analysis has usually been conducted by means of chemometric techniques and, to a lower extent, backpropagation artificial neural networks. Despite the excellent classification capacities of the latter, its development can be time-consuming and computer-intensive. Probabilistic artificial neural networks represent another attractive, yet scarcely evaluated classification technique which could be employed for the same purpose. The present work was aimed at comparing the performance of two chemometric techniques (principal component analysis and logistic regression) and two artificial neural networks (a backpropagation and a probabilistic neural network) as screening tools for cancer. The concentrations of copper, iron, selenium and zinc, which are known to play an important role in body chemistry, were used as indicators of physical status. Such elements were determined by total reflection X-ray fluorescence spectrometry in a sample of blood serum taken from individuals who had been given a positive ( N = 27) or a negative ( N = 32) diagnostic on various forms of cancer. The principal components analysis served two purposes: (i) initial screening of the data; and, (ii) reducing the dimension of the data space to be input to the networks. The logistic regression, as well as both artificial neural networks showed an outstanding performance in terms of distinguishing healthy (specificity: 90–100%) or unhealthy individuals (sensitivity: 100%). The probabilistic neural network offered additional advantages when compared to the more popular backpropagation counterpart. Effectively, the former is easier and faster to develop, for a smaller number of variables has to be optimized, and there are no risk of overtraining.

Automated Identification of Optically Sensed Aphid (Homoptera: Aphidae) Wingbeat Waveforms

An optical sensor was used to make digital recordings of wingbeat waveforms for the five most common aphids found on Guam: Aphis craccivora Koch, A. gossypii Glover, A. nerii Fonscolombe, Pentalonia nigronervosa Coquerel, and Toxoptera citricida (Kirkaldy). Wingbeat frequencies for each species overlapped all other species. However, mean wingbeat frequencies were significantly different for all species. Wingbeat frequencies and harmonic patterns were extracted from the recordings and submitted to cluster analysis, which failed to separate species completely. Several nearest neighbor and probabilistic neural network classifiers were built using time series, frequency spectra, wingbeat frequencies, and harmonic patterns as input variables. These classifiers were evaluated by having them identify wingbeat waveforms from aphids collected and recorded after their construction. The best performing classifier model was a probabilistic artificial neural network trained using 256-bin frequency spectra as input. Sixty-nine percent of the waveforms presented to this network were identified correctly. This study demonstrates the feasibility of developing an insect flight monitor that automatically counts and identifies individual flying insects. Essential components of the monitoring system are a photosensor, a multimedia personal computer, and software that identifies wingbeat frequency spectra using an artificial neural network.