Intrusion Detection System on KDD’99 Dataset with Imbalanced Classes

The paper describes a method of intrusion detection that keeps check of it with help of machine learning algorithms. The experiments have been conducted over KDD’99 cup dataset, which is an imbalanced dataset, cause of which recall of some classes coming drastically low as there were not enough instances of it in there. For Preprocessing of dataset One Hot Encoding and Label Encoding to make it machine readable. The dimensionality of dataset has been reduced using Principal Component Analysis and classification of dataset into classes viz. attack and normal is done by Naïve Bayes Classifier. Due to imbalanced nature, shift of focus was on recall and overall recall and compared with other models which have achieved great accuracy. Based on the results, using a self optimizing loop, model has achieved better geometric mean accuracy.

Enhanced location tracking in sensor fusion-assisted virtual reality micro-manipulation environments

Virtual reality (VR) technology plays a significant role in many biomedical applications. These VR scenarios increase the valuable experience of tasks requiring great accuracy with human subjects. Unfortunately, commercial VR controllers have large positioning errors in a micro-manipulation task. Here, we propose a VR-based framework along with a sensor fusion algorithm to improve the microposition tracking performance of a microsurgical tool. To the best of our knowledge, this is the first application of Kalman filter in a millimeter scale VR environment, by using the position data between the VR controller and an inertial measuring device. This study builds and tests two cases: (1) without sensor fusion tracking and (2) location tracking with active sensor fusion. The static and dynamic experiments demonstrate that the Kalman filter can provide greater precision during micro-manipulation in small scale VR scenarios.

A word-based account of comprehension and production of Kinyarwanda nouns in the Discriminative Lexicon

What are the cognitive units in the mental lexicon of Bantu speakers, words or morphemes? The very small experimental literature addressing this question suggests that the answer is morphemes, but a closer look at the results shows that this is answer is premature. A novel theory of the mental lexicon, the Discriminative Lexicon, which incorporates a word-based view of the mental lexicon and is computational implemented in the Linear Discriminative Learner (LDL), is put to the test with a data set of 11180 Kinyarwanda nouns. LDL is used to model comprehension and production of the nouns in the data set. LDL predicts comprehension and production of nouns with great accuracy. We conclude that the cognitive units in the mental lexicon of Kinyarwanda speakers are words.

DETECTION OF LUNG CANCER USING IMAGE PROCESSING

This project constructs and assesses an image processing approach for lung cancer diagnosis in this study. Image processing techniques are frequently utilized for picture improvement in the detection phase to enable early medical therapy in a variety of medical issues. We suggested a lung cancer detection approach based on picture segmentation in this study. Image segmentation is a level of image processing that is intermediate. To segment a CT scan image, a marker control watershed and region growth technique is applied. Following the detection phases, picture augmentation with the Gabor filter, image segmentation, and feature extraction is performed. We discovered the efficiency of our strategy based on the experimental results. The results demonstrate that the watershed with the masking method, which has great accuracy and robustness, is the best strategy for detecting major features. Keywords: Lung cancer, MATLAB, CT images, Distortion removal, Segmentation, Mortality rate.

DEVELOPMENT OF A MODEL FOR ASSESSMENT OF THE CONDITION OF PATIENTS WITH DISEASES OF THE SPINE BASED ON THE METHOD OF MULTI-CRITERIAL SELECTION OF ALTERNATIVES

Современная медицина имеет два приоритетных направления развития. Первое направление - это создание новейших лекарственных препаратов, а также разработка вакцин против новых вирусов. Второе направление - повсеместное и поэтапное внедрение в медицину информационных технологий. С болью в спине могут столкнуться люди не только пожилого возраста, но и подростки и даже грудные дети. Боль эта может быть вызвана многими причинами: как усталостью, так и всевозможными заболеваниями, которые могли развиться со временем или быть от рождения. По данным статистики ВОЗ, 80% населения страдает клиническими проявлениями остеохондроза позвоночника. В Российской Федерации большая часть амбулаторного приема неврологов и ортопедов занимают болезни позвоночника. Зачастую эксперту сложно однозначно оценить объект по некоторому критерию, возникают сомнения и поиски усредненной оценки. Но нередко затруднения в точном определении значения возникают не из-за недостатка опыта, а как раз, наоборот, из-за интуитивного понимания размытости оценки. Излишняя точность понятия может привести к потере части наилучших альтернатив или неправильному их ранжированию, если таковое применяется. Поэтому возникает необходимость разработки все более гибких по отношению к человеческому восприятию информации методов, позволяющих учитывать неопределенность все в большем количестве измерений. Целью данной работы является изучение методов, позволяющих с большой точностью определить заболевание позвоночника по некоторым жалобам пациента. Задачами работы являются выбор методов для просчета альтернатив, выбор критериев для альтернатив, и собственно, сами расчеты по выбранным методам Modern medicine has two priority areas of development. The first direction is the creation of the latest drugs, as well as the development of vaccines against new viruses. The second direction is the widespread and gradual introduction of information technologies into medicine. Back pain can be experienced not only by the elderly, but also by adolescents and even infants. This pain can be caused by many reasons: both fatigue and all kinds of diseases that could develop over time or be from birth. According to WHO statistics, 80% of the population suffers from clinical manifestations of osteochondrosis of the spine. In the Russian Federation, most of the outpatient visits to neurologists and orthopedists are spinal diseases. It is often difficult for an expert to unambiguously evaluate an object according to some criterion; doubts arise and searches for an average assessment. But often difficulties in accurately determining the meaning arise not because of a lack of experience, but, on the contrary, because of the intuitive understanding of the fuzziness of the assessment. Excessive precision of the concept can lead to the loss of some of the best alternatives or their incorrect ranking, if applicable. Therefore, there is a need to develop more and more flexible methods in relation to human perception of information, allowing to take into account the uncertainty in more and more dimensions. The aim of this work is to study methods that allow to determine with great accuracy the disease of the spine based on some of the patient's complaints. The tasks of the work are the choice of methods for calculating alternatives, the choice of criteria for alternatives, and, in fact, the calculations themselves according to the selected methods

Automatic Determination of Agricultural Plant Diseases

Machine learning technologies have developed rapidly in recent years, and people are now able to use them in various spheres of life, making their lives easier and better. The agro-industry is not lagging behind, and every year more and more problems in this area are solved with the help of machine learning algorithms. However, among the problems that have not yet been solved is the problem of identifying diseases of agricultural plants. According to the UN research, about 40% of the world’s harvest dies each year from various diseases, most of which could be avoided through timely intervention and treatment.To solve this problem, we offer an easy, accessible service for everyone, which will allow one to predict by the image of the plant leaves whether it is sick or healthy, or whether it needs any help or intrusion. This service will be indispensable for small farms engaged in growing crops. Thus, it will allow employees of such enterprises to immediately detect diseases and receive recommendations for the care of plants important to them.Therefore, it was decided to develop a neural network architecture that will solve this problem: the prediction of a plant disease by the image of its leaves. This neural network model is lightweight, does not take much time to learn, and has high accuracy on our dataset. It was also investigated which popular architectures (e.g. XceptionNet, DenseNet, etc.) of deep neural networks can have great accuracy in solving this problem. To realize the possibility of using the model by end users, i.e. farmers, it was decided to develop a special web service in the form of a telegram bot. With this bot, anyone can upload images of the leaves of agricultural plants and check whether this plant is healthy or free of any diseases. This bot is also trained to give appropriate advice to gardeners on the treatment of diseases or the proper cultivation of healthy plants.This solution fully solves the problem and has every chance to become an indispensable helper in preserving the world harvest.

FedVoting: A Cross-Silo Boosting Tree Construction Method for Privacy-Preserving Long-Term Human Mobility Prediction

The prediction of human mobility can facilitate resolving many kinds of urban problems, such as reducing traffic congestion, and promote commercial activities, such as targeted advertising. However, the requisite personal GPS data face privacy issues. Related organizations can only collect limited data and they experience difficulties in sharing them. These data are in “isolated islands” and cannot collectively contribute to improving the performance of applications. Thus, the method of federated learning (FL) can be adopted, in which multiple entities collaborate to train a collective model with their raw data stored locally and, therefore, not exchanged or transferred. However, to predict long-term human mobility, the performance and practicality would be impaired if only some models were simply combined with FL, due to the irregularity and complexity of long-term mobility data. Therefore, we explored the optimized construction method based on the high-efficient gradient-boosting decision tree (GBDT) model with FL and propose the novel federated voting (FedVoting) mechanism, which aggregates the ensemble of differential privacy (DP)-protected GBDTs by the multiple training, cross-validation and voting processes to generate the optimal model and can achieve both good performance and privacy protection. The experiments show the great accuracy in long-term predictions of special event attendance and point-of-interest visits. Compared with training the model independently for each silo (organization) and state-of-art baselines, the FedVoting method achieves a significant accuracy improvement, almost comparable to the centralized training, at a negligible expense of privacy exposure.

A NEUROPHYSIOLOGY-BASED MODEL TO ESTIMATE VISIBILITY IN ACTUAL LIGHTING ENVIRONMENT

Visibility is one of the fundamental factors for explaining the lighting environment, and various quantitative indexes for these have been made in such a manner that directly finds the formulae that best match the numerous subjective evaluations. Recently neurophysiology-based models have been paid attention as the complementary or alternative methods to these conventional indexes. In this paper we will introduce the normalization and gain control models, which are generally accepted theories in the visual information processing, to estimate the visibility, and verify the validity of this algorithm by comparing with the subjective evaluation in the experimental room. The result showed that the algorithm in this research could estimate the visibility of the simple objects in the experimental space with a pseudo window with great accuracy, however, it is desirable and necessary to publicly discuss various algorithms and clarify their reliability based on the verification results under various situations.

Magnetoelastic Ribbons as Vibration Sensors for Real-Time Health Monitoring of Rotating Metal Beams

In the current work, magnetoelastic material ribbons are used as vibration sensors to monitor, in real time and non-destructively, the mechanical health state of rotating beam blades. The magnetoelastic material has the form of a thin ribbon and is composed of Metglas alloy 2826 MB. The study was conducted in two stages. In the first stage, an experiment was performed to test the ability of the ribbon to detect and transmit the vibration behavior of four rotating blades, while the second stage was the same as the first but with minor damages introduced to the blades. As far as the first stage is concerned, the results show that the sensor can detect and transmit with great accuracy the vibratory behavior of the rotating blades, through which important information about the mechanical health state of the blade can be extracted. Specifically, the fast Fourier transform (FFT) spectrum of the recorded signal revealed five dominant peaks in the frequency range 0–3 kHz, corresponding to the first five bending modes of the blades. The identification process was accomplished using ANSYS modal analysis, and the comparison results showed deviation values of less than 1% between ANSYS and the experimental values. In the second stage, two types of damages were introduced to the rotating blades, an edge cut and a hole. The damages were scaled in number from one blade to another, with the first blade having only one side cut while the last blade had two side cuts and two holes. The results, as was expected, show a measurable shifting on the frequency values of the bending modes, thus proving the ability of the proposed magnetoelastic sensors to detect and transmit changes of the mechanical state of rotating blades in real time.

Two approaches to quantum gravity and M-(atrix) theory at large number of dimensions

A Gaussian approximation to the bosonic part of M-(atrix) theory with mass deformation is considered at large values of the dimension d. From the perspective of the gauge/gravity duality this action reproduces with great accuracy the stringy Hagedorn phase transition from a confinement (black string) phase to a deconfinement (black hole) phase whereas from the perspective of the matrix/geometry approach this action only captures a remnant of the geometric Yang–Mills-to-fuzzy-sphere phase where the fuzzy sphere solution is only manifested as a three-cut configuration termed the “baby fuzzy sphere” configuration. The Yang–Mills phase retains most of its characteristics with two exceptions: (i) the uniform distribution inside a solid ball suffers a crossover at very small values of the gauge coupling constant to a Wigner’s semicircle law, and (ii) the uniform distribution at small values of the temperatures is nonexistent.