Search results for: random number

Authors: Songul Cinaroglu

Abstract:

Background: Random Forest is an efficient, multi-class machine learning method using for classification, regression and other tasks. This method is operating by constructing each tree using different bootstrap sample of the data. Determining the number of trees in random forests is an open question in the literature for studies about improving classification performance of random forests. Aim: The aim of this study is to analyze whether there is an optimal number of trees in Random Forests and how performance of Random Forests differ according to increase in number of trees using sample health data sets in R programme. Method: In this study we analyzed the performance of Random Forests as the number of trees grows and doubling the number of trees at every iteration using “random forest” package in R programme. For determining minimum and optimal number of trees we performed Mc Nemar test and Area Under ROC Curve respectively. Results: At the end of the analysis it was found that as the number of trees grows, it does not always means that the performance of the forest is better than forests which have fever trees. In other words larger number of trees only increases computational costs but not increases performance results. Conclusion: Despite general practice in using random forests is to generate large number of trees for having high performance results, this study shows that increasing number of trees doesn’t always improves performance. Future studies can compare different kinds of data sets and different performance measures to test whether Random Forest performance results change as number of trees increase or not.

Keywords: classification methods, decision trees, number of trees, random forest

Procedia PDF Downloads 396

Authors: Melvin Ballera, Aldrich Olivar, Mary Soriano

Abstract:

Digital computers nowadays must be able to have a utility that is capable of generating random numbers. Usually, computer-generated random numbers are not random given predefined values such as starting point and end points, making the sequence almost predictable. There are many applications of random numbers such business simulation, manufacturing, services domain, entertainment sector and other equally areas making worthwhile to design a unique method and to allow unpredictable random numbers. Applying stochastic simulation using linear congruential algorithm, it shows that as it increases the numbers of the seed and range the number randomly produced or selected by the computer becomes unique. If this implemented in an environment where random numbers are very much needed, the reliability of the random number is guaranteed.

Keywords: stochastic simulation, random numbers, linear congruential algorithm, pseudorandomness

Procedia PDF Downloads 318

Authors: M. Hamdi, R. Rhouma, S. Belghith

Abstract:

Generating random numbers are mainly used to create secret keys or random sequences. It can be carried out by various techniques. In this paper we present a very simple and efficient pseudo-random number generator (PRNG) based on chaotic maps and S-Box tables. This technique adopted two main operations one to generate chaotic values using two logistic maps and the second to transform them into binary words using random S-Box tables. The simulation analysis indicates that our PRNG possessing excellent statistical and cryptographic properties.

Keywords: Random Numbers, Chaotic map, S-box, cryptography, statistical tests

Procedia PDF Downloads 366

Authors: Roman Snytsar

Abstract:

Pseudo-random numbers are often used in scientific computing such as the Monte Carlo Simulations or the Quantum Inspired Optimization. Requirements for a parallel random number generator running in the modern multi-core vector environment are more stringent than those for sequential random number generators. As well as passing the usual quality tests, the output of the parallel random number generator must be verifiable and reproducible throughout the concurrent execution. We propose a family of vectorized Permuted Congruential Generators. Implementations are available for multiple modern vector modern computer architectures. Besides demonstrating good single core performance, the generators scale easily across many processor cores and multiple distributed nodes. We provide performance and parallel speedup analysis and comparisons between the implementations.

Keywords: pseudo-random numbers, quantum optimization, SIMD, parallel computing

Procedia PDF Downloads 120

Authors: S.Balaji, R.Sujatha, M. Ramakrishnan

Abstract:

Signcryption is a cryptographic primitive that fulfills both the functions of digital signature and public key encryption simultaneously in low cost when compared with the traditional signature-then-encryption approach. In this paper, we propose a novel mouse movement based key generation technique to generate secret keys which is secure against the outer and insider attacks. Tag Key Encapsulation Mechanism (KEM) process is implemented using True Random Number Generator (TRNG) method. This TRNG based key is used for data encryption in the Data Encapsulation Mechanism (DEM). We compare the statistical reports of the proposed system with the previous methods which implements TKEM based on pseudo random number generator

Keywords: pseudo random umber generator, signcryption, true random number generator, node deployment

Procedia PDF Downloads 343

Authors: Hicham Salhi, Mohamed Si-Ameur, Nadjib Chafai

Abstract:

Natural convection of a nanofluid consisting of water and nanoparticles (Ag or TiO2) in an inclined enclosure cavity, has been studied numerically, heated by a (random temperature, based on the random function). The governing equations are solved numerically using the finite-volume. Results are presented in the form of streamlines, isotherms, and average Nusselt number. In addition, a parametric study is carried out to examine explicitly the volume fraction effects of nanoparticles (Ψ= 0.1, 0.2), the Rayleigh number (Ra=103, 104, 105, 106),the inclination angle of the cavity( égale à 0°, 30°, 45°, 90°, 135°, 180°), types of temperature (constant ,random), types of (NF) (Ag andTiO2). The results reveal that (NPs) addition remarkably enhances heat transfer in the cavity especially for (Ψ= 0.2). Besides, the effect of inclination angle and type of temperature is more pronounced at higher Rayleigh number.

Keywords: nanofluid, natural convection, inclined cavity, random temperature, finite-volume

Procedia PDF Downloads 289

Authors: Mochamad Beta Auditama, Yusuf Kurniawan

Abstract:

A smartphone or tablet require a strong randomness to establish secure encrypted communication, encrypt files, etc. Therefore, random number generation is one of the main keys to provide secrecy. Android devices are equipped with hardware-based sensors, such as accelerometer, gyroscope, etc. Each of these sensors provides a stochastic process which has a potential to be used as an extra randomness source, in addition to /dev/random and /dev/urandom pseudorandom number generators. Android sensors can provide randomness automatically. To obtain randomness from Android sensors, each one of Android sensors shall be used to construct an entropy source. After all entropy sources are constructed, output from these entropy sources are combined to provide more entropy. Then, a deterministic process is used to produces a sequence of random bits from the combined output. All of these processes are done in accordance with NIST SP 800-22 and the series of NIST SP 800-90. The operation conditions are done 1) on Android user-space, and 2) the Android device is placed motionless on a desk.

Keywords: Android hardware-based sensor, deterministic process, entropy source, random number generation/generators

Procedia PDF Downloads 376

Authors: Sai Charan Kamana, Harsha Vardhan Nakkina, B.R. Chandavarkar

Abstract:

In a highly secure and robust key generation process, a key role is played by randomness and random numbers when current real-world cryptosystems are observed. Most of the present-day cryptographic protocols depend upon the Random Number Generators (RNG), Pseudo-Random Number Generator (PRNG). These protocols often use noisy channels such as Disk seek time, CPU temperature, Mouse pointer movement, Fan noise to obtain true random values. Despite being cost-effective, these noisy channels may need additional hardware devices to continuously communicate with them. On the other hand, Hash functions are Pseudo-Random (because of their requirements). So, they are a good replacement for these noisy channels and have low hardware requirements. This paper discusses, some of the key generation methodologies, and their drawbacks. This paper explains how hash functions can be used in key generation, how to combine Key Derivation Functions with hash functions.

Keywords: key derivation, hash based key derivation, password based key derivation, symmetric key derivation

Procedia PDF Downloads 162

Authors: Somaiyeh Dehghan, Mohammad Reza Kheirkhahan Haghighi

Abstract:

The rapid growth of domains that have data with a large number of features, while the number of samples is limited has caused difficulty in constructing strong classifiers. To reduce the dimensionality of the feature space becomes an essential step in classification task. Random subspace method (or attribute bagging) is an ensemble classifier that consists of several classifiers that each base learner in ensemble has subset of features. In the present paper, we introduce Random Subspace Ensemble of CMAC neural network (RSE-CMAC), each of which has training with subset of features. Then we use this model for classification task. For evaluation performance of our model, we compare it with bagging algorithm on 36 UCI datasets. The results reveal that the new model has better performance.

Keywords: classification, random subspace, ensemble, CMAC neural network

Procedia PDF Downloads 332

Authors: Kevin Worrall, Euan McGookin

Abstract:

The use of optimization algorithms is common across a large number of diverse fields. This work presents the use of a hybrid optimization algorithm applied to a mobile robot tasked with carrying out a search of an unknown environment. The algorithm is then applied to the multiple robots case, which results in a reduction in the time taken to carry out the search. The hybrid algorithm is a Random Search Algorithm fused with a Tabu mechanism. The work shows that the algorithm locates the desired points in a quicker time than a brute force search. The Tabu Random algorithm is shown to work within a simulated environment using a validated mathematical model. The simulation was run using three different environments with varying numbers of targets. As an algorithm, the Tabu Random is small, clear and can be implemented with minimal resources. The power of the algorithm is the speed at which it locates points of interest and the robustness to the number of robots involved. The number of robots can vary with no changes to the algorithm resulting in a flexible algorithm.

Keywords: algorithms, control, multi-agent, search and rescue

Procedia PDF Downloads 240

Authors: D. S. Palimkar

Abstract:

The FRIGDI (functional random integrodifferential inclusion) seems to be new and includes several known random differential inclusions already studied in the literature as special cases have been discussed in the literature for various aspects of the solutions. In this paper, we prove the existence result for FIGDI under the non-convex case of multi-valued function involved in it.Using random fixed point theorem of B. C. Dhage and caratheodory condition. This result is new to the theory of differential inclusion.

Keywords: caratheodory condition, random differential inclusion, random solution, integro-differential inclusion

Procedia PDF Downloads 466

Authors: Rajkumar N. Ingle

Abstract:

In this paper, the existence of a solution of nonlinear functional random differential equations of the first order is proved under caratheodory condition. The study of the functional random differential equation has got importance in the random analysis of the dynamical systems of universal phenomena. Objectives: Nonlinear functional random differential equation is useful to the scientists, engineers, and mathematicians, who are engaged in N.F.R.D.E. analyzing a universal random phenomenon, govern by nonlinear random initial value problems of D.E. Applications of this in the theory of diffusion or heat conduction. Methodology: Using the concepts of probability theory, functional analysis, generally the existence theorems for the nonlinear F.R.D.E. are prove by using some tools such as fixed point theorem. The significance of the study: Our contribution will be the generalization of some well-known results in the theory of Nonlinear F.R.D.E.s. Further, it seems that our study will be useful to scientist, engineers, economists and mathematicians in their endeavors to analyses the nonlinear random problems of the universe in a better way.

Keywords: Random Fixed Point Theorem, functional random differential equation, N.F.R.D.E., universal random phenomenon

Procedia PDF Downloads 503

Authors: M. Khalid, G. N. Singh

Abstract:

In this paper, we have considered the problems of estimation for the population mean on current (second) occasion in two-occasion successive sampling under random non-response situations. Some modified exponential type estimators have been proposed and their properties are studied under the assumptions that the number of sampling unit follows a discrete distribution due to random non-response situations. The performances of the proposed estimators are compared with linear combinations of two estimators, (a) sample mean estimator for fresh sample and (b) ratio estimator for matched sample under the complete response situations. Results are demonstrated through empirical studies which present the effectiveness of the proposed estimators. Suitable recommendations have been made to the survey practitioners.

Keywords: modified exponential estimator, successive sampling, random non-response, auxiliary variable, bias, mean square error

Procedia PDF Downloads 349

Authors: Kamaljit Pati, Manas Kumar Mohanty, Sanjib Sadhu

Abstract:

A heuristic has been designed to generate a random simple monotone polygon from a given set of ‘n’ points lying on a 2-Dimensional plane. Our heuristic generates a random monotone polygon in O(n) time after O(nℓogn) preprocessing time which is improved over the previous work where a random monotone polygon is produced in the same O(n) time but the preprocessing time is O(k) for n < k < n2. However, our heuristic does not generate all possible random polygons with uniform probability. The space complexity of our proposed heuristic is O(n).

Keywords: sorting, monotone polygon, visibility, chain

Procedia PDF Downloads 428

Authors: Mustapha Benssalah, Mustapha Djeddou, Karim Drouiche

Abstract:

The random number generation (RNG) presents a significant importance for the security and the privacy of numerous applications, such as RFID technology and smart cards. Since, the quality of the generated bit sequences is paramount that a weak internal generator for example, can directly cause the entire application to be insecure, and thus it makes no sense to employ strong algorithms for the application. In this paper, we propose a new pseudo random number generator (PRNG), suitable for cryptosystems ECC-based, constructed by randomly selecting points from several elliptic curves randomly selected. The main contribution of this work is the increasing of the generator internal states by extending the set of its output realizations to several curves auto-selected. The quality and the statistical characteristics of the proposed PRNG are validated using the Chi-square goodness of fit test and the empirical Special Publication 800-22 statistical test suite issued by NIST.

Keywords: PRNG, security, cryptosystem, ECC

Procedia PDF Downloads 445

Authors: Ali Nourollah, Mohsen Movahedinejad

Abstract:

In this paper a new algorithm to generate random simple polygons from a given set of points in a two dimensional plane is designed. The proposed algorithm uses a genetic algorithm to generate polygons with few vertices. A new merge algorithm is presented which converts any two polygons into a simple polygon. This algorithm at first changes two polygons into a polygonal chain and then the polygonal chain is converted into a simple polygon. The process of converting a polygonal chain into a simple polygon is based on the removal of intersecting edges. The merge algorithm has the time complexity of O ((r+s) *l) where r and s are the size of merging polygons and l shows the number of intersecting edges removed from the polygonal chain. It will be shown that 1 < l < r+s. The experiments results show that the proposed algorithm has the ability to generate a great number of different simple polygons and has better performance in comparison to celebrated algorithms such as space partitioning and steady growth.

Keywords: Divide and conquer, genetic algorithm, merge polygons, Random simple polygon generation.

Procedia PDF Downloads 534

Authors: Weisi Guo

Abstract:

It is well established that there may be running correlations in sequential exam marks due to students sitting in the order of course registration patterns. As such, a random and non-sequential sampling of exam marks is a standard recommended practice. Here, the paper examines a large number of exam data stretching several years across different modules to see the degree to which it is true. Using the real mark distribution as a generative process, it was found that random simulated data had no more sequential randomness than the real data. That is to say, the running correlations that one often observes are statistically identical to chance. Digging deeper, it was found that some high running correlations have students that indeed share a common course history and make similar mistakes. However, at the statistical scale of a module question, the combined effect is statistically similar to the random shuffling of papers. As such, there may not be the need to take random samples for marks, but it still remains good practice to mark papers in a random sequence to reduce the repetitive marking bias and errors.

Keywords: data analysis, empirical study, exams, marking

Procedia PDF Downloads 183

Authors: Adi A. Maaita, Hamza A. A. Al Sewadi

Abstract:

One of the crucial parameters of digital cryptographic systems is the selection of the keys used and their distribution. The randomness of the keys has a strong impact on the system’s security strength being difficult to be predicted, guessed, reproduced or discovered by a cryptanalyst. Therefore, adequate key randomness generation is still sought for the benefit of stronger cryptosystems. This paper suggests an algorithm designed to generate and test pseudo random number sequences intended for cryptographic applications. This algorithm is based on mathematically manipulating a publically agreed upon information between sender and receiver over a public channel. This information is used as a seed for performing some mathematical functions in order to generate a sequence of pseudorandom numbers that will be used for encryption/decryption purposes. This manipulation involves permutations and substitutions that fulfills Shannon’s principle of “confusion and diffusion”. ASCII code characters wereutilized in the generation process instead of using bit strings initially, which adds more flexibility in testing different seed values. Finally, the obtained results would indicate sound difficulty of guessing keys by attackers.

Keywords: cryptosystems, information security agreement, key distribution, random numbers

Procedia PDF Downloads 269

Authors: John White

Abstract:

This study investigated the effects of practice schedules on learning and performance in golf putting, specifically focusing on the impact of increasing contextual interference (CI). University students (n=7) were randomly assigned to blocked, random, or increasing practice schedules. During acquisition, participants performed 135 putting trials using different weighted golf balls. The blocked group followed a specific sequence of ball weights, while the random group practiced with the balls in a random order. The increasing group started with a blocked schedule, transitioned to a serial schedule, and concluded with a random schedule. Retention and transfer tests were conducted 24 hours later. The results indicated that high levels of CI (random practice) were more beneficial for learning than low levels of CI (blocked practice). The increasing practice schedule, incorporating blocked, serial, and random practice, demonstrated advantages over traditional blocked and random schedules. Additionally, EEG was used to explore the neurophysiological effects of the increasing practice schedule.

Keywords: skill acquisition, motor control, learning, contextual interference

Procedia PDF Downloads 96

Authors: Michael A. Kraus, Navid Pourmoghaddam, Martin Botz, Jens Schneider, Geralt Siebert

Abstract:

Fragmentation analysis of tempered glass gives insight into the quality of the tempering process and defines a certain degree of safety as well. Different standard such as the European EN 12150-1 or the American ASTM C 1048/CPSC 16 CFR 1201 define a minimum number of fragments required for soda-lime safety glass on the basis of fragmentation test results for classification. This work presents an approach for the glass breakage pattern prediction using a Voronoi Tesselation over Random Fields. The random Voronoi tessellation is trained with and validated against data from several breakage patterns. The fragments in observation areas of 50 mm x 50 mm were used for training and validation. All glass specimen used in this study were commercially available soda-lime glasses at three different thicknesses levels of 4 mm, 8 mm and 12 mm. The results of this work form a Bayesian framework for the training and prediction of breakage patterns of tempered soda-lime glass using a Voronoi Random Field Tesselation. Uncertainties occurring in this process can be well quantified, and several statistical measures of the pattern can be preservation with this method. Within this work it was found, that different Random Fields as basis for the Voronoi Tesselation lead to differently well fitted statistical properties of the glass breakage patterns. As the methodology is derived and kept general, the framework could be also applied to other random tesselations and crack pattern modelling purposes.

Keywords: glass breakage predicition, Voronoi Random Field Tessellation, fragmentation analysis, Bayesian parameter identification

Procedia PDF Downloads 161

Authors: Samir Benzegane, Said Sadoudi, Mustapha Djeddou

Abstract:

In this paper, we present a software development of video streaming encryption for Device-to-Device (D2D) communications by using Hyperchaos-based Random Number Generator (HRNG) implemented in C#. The software implements and uses the proposed HRNG to generate key stream for encrypting and decrypting real-time video data. The used HRNG consists of Hyperchaos Lorenz system which produces four signal outputs taken as encryption keys. The generated keys are characterized by high quality randomness which is confirmed by passing standard NIST statistical tests. Security analysis of the proposed encryption scheme confirms its robustness against different attacks.

Keywords: hyperchaos Lorenz system, hyperchaos-based random number generator, D2D communications, C#

Procedia PDF Downloads 374

Authors: D. S. Palimkar

Abstract:

Random fixed point theory has received much attention in recent years, and it is needed for the study of various classes of random equations. The study of random fixed point theorems was initiated by the Prague school of probabilistic in the 1950s. The existence and uniqueness of fixed points for the self-maps of a metric space by altering distances between the points with the use of a control function is an interesting aspect in the classical fixed point theory. In a new category of fixed point problems for a single self-map with the help of a control function that alters the distance between two points in a metric space which they called an altering distance function. In this paper, we prove the results of existence of random common fixed point and its uniqueness for a pair of random mappings under weakly contractive condition for generalizing alter distance function in polish spaces using Random Common Fixed Point Theorem for Generalized Weakly Contractions.

Keywords: Polish space, random common fixed point theorem, weakly contractive mapping, altering function

Procedia PDF Downloads 275

Authors: Y. B. Kirankumar, J. D. Mallapur

Abstract:

Wireless Sensor Network (WSN) is an emerging technology, which has great invention for various low cost applications both for mass public as well as for defence. The wireless sensor communication technology allows random participation of sensor nodes with particular applications to take part in the network, which results in most of the uncovered simulation area, where fewer nodes are located at far distances. The drawback of such network would be that the additional energy is spent by the nodes located in a pattern of dense location, using more number of nodes for a smaller distance of communication adversely in a region with less number of nodes and additional energy is again spent by the source node in order to transmit a packet to neighbours, thereby transmitting the packet to reach the destination. The proposed work is intended to develop Energy Efficient Node Placement Algorithm (EENPA) in order to place the sensor node efficiently in simulated area, where all the nodes are equally located on a radial path to cover maximum area at equidistance. The total energy consumed by each node compared to random placement of nodes is less by having equal burden on fewer nodes of far location, having distributed the nodes in whole of the simulation area. Calculating the network lifetime also proves to be efficient as compared to random placement of nodes, hence increasing the network lifetime, too. Simulation is been carried out in a qualnet simulator, results are obtained on par with random placement of nodes with EENP algorithm.

Keywords: energy, WSN, wireless sensor network, energy approach

Procedia PDF Downloads 313

Authors: Kazuma Okada, Tomoaki Hashimoto, Hirokazu Tahara

Abstract:

Recently, the effectiveness of random dither quantization method for linear feedback control systems has been shown in several papers. However, the random dither quantization method has not yet been applied to nonlinear feedback control systems. The objective of this paper is to verify the effectiveness of random dither quantization method for nonlinear feedback control systems. For this purpose, we consider the attitude stabilization problem of satellites using discrete-level actuators. Namely, this paper provides a control method based on the random dither quantization method for stabilizing the attitude of satellites using discrete-level actuators.

Keywords: quantized control, nonlinear systems, random dither quantization

Procedia PDF Downloads 243

Authors: Halil Ibrahim Demir, Alper Goksu, Onur Canpolat, Caner Erden, Melek Nur

Abstract:

Traditionally process planning, scheduling and due date assignment are performed sequentially and separately. High interrelation between these functions makes integration very useful. Although there are numerous works on integrated process planning and scheduling and many works on scheduling with due date assignment, there are only a few works on the integration of these three functions. Here we tested the different integration levels of these three functions and found a fully integrated version as the best. We applied genetic search and random search and genetic search was found better compared to the random search. We penalized all earliness, tardiness and due date related costs. Since all these three terms are all undesired, it is better to penalize all of them.

Keywords: process planning, scheduling, due-date assignment, genetic algorithm, random search

Procedia PDF Downloads 375

Authors: Wenjie Qin, Kewei Lv

Abstract:

In 1996, Boneh and Venkatesan proposed the Hidden Number Problem (HNP) and proved the most significant bits (MSB) of computational Diffie-Hellman key exchange scheme and related schemes are unpredictable bits. They also gave a method which is a lattice rounding technique to solve HNP in non-uniform model. In this paper, we put forward a new concept that is Schnorr-MSB-HNP. We also reduce the problem of solving Schnorr signature private key with a few consecutive most significant bits of random nonce (used at each signature generation) to Schnorr-MSB-HNP, then we use the rounding technique to solve the Schnorr-MSB-HNP. We have come to the conclusion that if there is a ‘miraculous box’ which inputs the random nonce and outputs 2loglogq (q is a prime number) most significant bits of nonce, the signature private key will be obtained by choosing 2logq signature messages randomly. Thus we get an attack on the Schnorr signature private key.

Keywords: rounding technique, most significant bits, Schnorr signature algorithm, nonce, Schnorr-MSB-HNP

Procedia PDF Downloads 234

Authors: Alhusein Almahjoub, Dongyu Qiu

Abstract:

This paper deals with the random access procedure in next-generation networks and presents the solution to reduce total service time (TST) which is one of the most important performance metrics in current and future internet of things (IoT) based networks. The proposed solution focuses on the calculation of optimal transmission probability which maximizes the success probability and reduces TST. It uses the information of several idle preambles in every time slot, and based on it, it estimates the number of backlogged IoT devices using Naïve Bayes estimation which is a type of supervised learning in the machine learning domain. The estimation of backlogged devices is necessary since optimal transmission probability depends on it and the eNodeB does not have information about it. The simulations are carried out in MATLAB which verify that the proposed solution gives excellent performance.

Keywords: random access, LTE/LTE-A, 5G, machine learning, Naïve Bayes estimation

Procedia PDF Downloads 146

Authors: Karima Kimouche

Abstract:

In this paper, we consider the asymptotic problems in spectral analysis of stationary causal random fields. We impose conditions only involving (conditional) moments, which are easily verifiable for a variety of nonlinear random fields. Limiting distributions of periodograms and smoothed periodogram spectral density estimates are obtained and applications to the spectral domain bootstrap are given.

Keywords: spatial nonlinear processes, spectral estimators, GMC condition, bootstrap method

Procedia PDF Downloads 455

Authors: Ja-Keoung Koo, Kensuke Nakamura, Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Byung-Woo Hong

Abstract:

The machine learning techniques based on a convolutional neural network (CNN) have been actively developed and successfully applied to a variety of image analysis tasks including reconstruction, noise reduction, resolution enhancement, segmentation, motion estimation, object recognition. The classical visual information processing that ranges from low level tasks to high level ones has been widely developed in the deep learning framework. It is generally considered as a challenging problem to derive visual interpretation from high dimensional imagery data. A CNN is a class of feed-forward artificial neural network that usually consists of deep layers the connections of which are established by a series of non-linear operations. The CNN architecture is known to be shift invariant due to its shared weights and translation invariance characteristics. However, it is often computationally intractable to optimize the network in particular with a large number of convolution layers due to a large number of unknowns to be optimized with respect to the training set that is generally required to be large enough to effectively generalize the model under consideration. It is also necessary to limit the size of convolution kernels due to the computational expense despite of the recent development of effective parallel processing machinery, which leads to the use of the constantly small size of the convolution kernels throughout the deep CNN architecture. However, it is often desired to consider different scales in the analysis of visual features at different layers in the network. Thus, we propose a CNN model where different sizes of the convolution kernels are applied at each layer based on the random projection. We apply random filters with varying sizes and associate the filter responses with scalar weights that correspond to the standard deviation of the random filters. We are allowed to use large number of random filters with the cost of one scalar unknown for each filter. The computational cost in the back-propagation procedure does not increase with the larger size of the filters even though the additional computational cost is required in the computation of convolution in the feed-forward procedure. The use of random kernels with varying sizes allows to effectively analyze image features at multiple scales leading to a better generalization. The robustness and effectiveness of the proposed CNN based on random kernels are demonstrated by numerical experiments where the quantitative comparison of the well-known CNN architectures and our models that simply replace the convolution kernels with the random filters is performed. The experimental results indicate that our model achieves better performance with less number of unknown weights. The proposed algorithm has a high potential in the application of a variety of visual tasks based on the CNN framework. Acknowledgement—This work was supported by the MISP (Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by IITP, and NRF-2014R1A2A1A11051941, NRF2017R1A2B4006023.

Keywords: deep learning, convolutional neural network, random kernel, random projection, dimensionality reduction, object recognition

Procedia PDF Downloads 291

Authors: Songul Cinaroglu

Abstract:

Machine learning regression methods are recommended as an alternative to classical regression methods in the existence of variables which are difficult to model. Data for health expenditure is typically non-normal and have a heavily skewed distribution. This study aims to compare machine learning regression methods by hyperparameter tuning to predict health expenditure per capita. A multiple regression model was conducted and performance results of Lasso Regression, Random Forest Regression and Support Vector Machine Regression recorded when different hyperparameters are assigned. Lambda (λ) value for Lasso Regression, number of trees for Random Forest Regression, epsilon (ε) value for Support Vector Regression was determined as hyperparameters. Study results performed by using 'k' fold cross validation changed from 5 to 50, indicate the difference between machine learning regression results in terms of R², RMSE and MAE values that are statistically significant (p < 0.001). Study results reveal that Random Forest Regression (R² ˃ 0.7500, RMSE ≤ 0.6000 ve MAE ≤ 0.4000) outperforms other machine learning regression methods. It is highly advisable to use machine learning regression methods for modelling health expenditures.

Keywords: machine learning, lasso regression, random forest regression, support vector regression, hyperparameter tuning, health expenditure

Procedia PDF Downloads 226