Search results for: Conjugate Dirichlet kernel

Authors: Baljit Singh Khehra, Amar Partap Singh Pharwaha

Abstract:

Clusters of Microcalcifications (MCCs) are most frequent symptoms of Ductal Carcinoma in Situ (DCIS) recognized by mammography. Least-Square Support Vector Machine (LS-SVM) is a variant of the standard SVM. In the paper, LS-SVM is proposed as a classifier for classifying MCCs as benign or malignant based on relevant extracted features from enhanced mammogram. To establish the credibility of LS-SVM classifier for classifying MCCs, a comparative evaluation of the relative performance of LS-SVM classifier for different kernel functions is made. For comparative evaluation, confusion matrix and ROC analysis are used. Experiments are performed on data extracted from mammogram images of DDSM database. A total of 380 suspicious areas are collected, which contain 235 malignant and 145 benign samples, from mammogram images of DDSM database. A set of 50 features is calculated for each suspicious area. After this, an optimal subset of 23 most suitable features is selected from 50 features by Particle Swarm Optimization (PSO). The results of proposed study are quite promising.

Keywords: Clusters of Microcalcifications, Ductal Carcinoma in Situ, Least-Square Support Vector Machine, Particle Swarm Optimization.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1754

Authors: Wei-Jong Yang, Yu-Siang Su, Pau-Choo Chung, Jar-Ferr Yang

Abstract:

Moving object detection (MOD) is an important issue in advanced driver assistance systems (ADAS). There are two important moving objects, pedestrians and scooters in ADAS. In real-world systems, there exist two important challenges for MOD, including the computational complexity and the detection accuracy. The histogram of oriented gradient (HOG) features can easily detect the edge of object without invariance to changes in illumination and shadowing. However, to reduce the execution time for real-time systems, the image size should be down sampled which would lead the outlier influence to increase. For this reason, we propose the histogram of uniformly-oriented gradient (HUG) features to get better accurate description of the contour of human body. In the testing phase, the support vector machine (SVM) with linear kernel function is involved. Experimental results show the correctness and effectiveness of the proposed method. With SVM classifiers, the real testing results show the proposed HUG features achieve better than classification performance than the HOG ones.

Keywords: Moving object detection, histogram of oriented gradient histogram of oriented gradient, histogram of uniformly-oriented gradient, linear support vector machine.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1181

Authors: Marina Velikova, Hennie Daniels, Ad Feelders

Abstract:

In many data mining applications, it is a priori known that the target function should satisfy certain constraints imposed by, for example, economic theory or a human-decision maker. In this paper we consider partially monotone prediction problems, where the target variable depends monotonically on some of the input variables but not on all. We propose a novel method to construct prediction models, where monotone dependences with respect to some of the input variables are preserved by virtue of construction. Our method belongs to the class of mixture models. The basic idea is to convolute monotone neural networks with weight (kernel) functions to make predictions. By using simulation and real case studies, we demonstrate the application of our method. To obtain sound assessment for the performance of our approach, we use standard neural networks with weight decay and partially monotone linear models as benchmark methods for comparison. The results show that our approach outperforms partially monotone linear models in terms of accuracy. Furthermore, the incorporation of partial monotonicity constraints not only leads to models that are in accordance with the decision maker's expertise, but also reduces considerably the model variance in comparison to standard neural networks with weight decay.

Keywords: mixture models, monotone neural networks, partially monotone models, partially monotone problems.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1193

Authors: Mahtab Davari, Charles Edward Okon, Somayeh Aghanavesi

Abstract:

Every institute of learning is usually interested in the performance of enrolled students. The level of these performances determines the approach an institute of study may adopt in rendering academic services. The focus of this paper is to evaluate students' academic performance in given courses of study using machine learning methods. This study evaluated various supervised machine learning classification algorithms such as Logistic Regression (LR), Support Vector Machine (SVM), Random Forest, Decision Tree, K-Nearest Neighbors, Linear Discriminant Analysis (LDA), and Quadratic Discriminant Analysis, using selected features to predict study performance. The accuracy, precision, recall, and F1 score obtained from a 5-Fold Cross-Validation were used to determine the best classification algorithm to predict students’ performances. SVM (using a linear kernel), LDA, and LR were identified as the best-performing machine learning methods. Also, using the LR model, this study identified students' educational habits such as reading and paying attention in class as strong determinants for a student to have an above-average performance. Other important features include the academic history of the student and work. Demographic factors such as age, gender, high school graduation, etc., had no significant effect on a student's performance.

Keywords: Student performance, supervised machine learning, prediction, classification, cross-validation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 461

Authors: Walaa Khalaf, Calogero Pace, Manlio Gaudioso

Abstract:

We present an Electronic Nose (ENose), which is aimed at identifying the presence of one out of two gases, possibly detecting the presence of a mixture of the two. Estimation of the concentrations of the components is also performed for a volatile organic compound (VOC) constituted by methanol and acetone, for the ranges 40-400 and 22-220 ppm (parts-per-million), respectively. Our system contains 8 sensors, 5 of them being gas sensors (of the class TGS from FIGARO USA, INC., whose sensing element is a tin dioxide (SnO2) semiconductor), the remaining being a temperature sensor (LM35 from National Semiconductor Corporation), a humidity sensor (HIH–3610 from Honeywell), and a pressure sensor (XFAM from Fujikura Ltd.). Our integrated hardware–software system uses some machine learning principles and least square regression principle to identify at first a new gas sample, or a mixture, and then to estimate the concentrations. In particular we adopt a training model using the Support Vector Machine (SVM) approach with linear kernel to teach the system how discriminate among different gases. Then we apply another training model using the least square regression, to predict the concentrations. The experimental results demonstrate that the proposed multiclassification and regression scheme is effective in the identification of the tested VOCs of methanol and acetone with 96.61% correctness. The concentration prediction is obtained with 0.979 and 0.964 correlation coefficient for the predicted versus real concentrations of methanol and acetone, respectively.

Keywords: Electronic nose, Least square regression, Mixture ofgases, Support Vector Machine.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2493

Authors: К. R. Aida–Zade, C. Ardil, S. S. Rustamov

Abstract:

Statement of the automatic speech recognition problem, the assignment of speech recognition and the application fields are shown in the paper. At the same time as Azerbaijan speech, the establishment principles of speech recognition system and the problems arising in the system are investigated. The computing algorithms of speech features, being the main part of speech recognition system, are analyzed. From this point of view, the determination algorithms of Mel Frequency Cepstral Coefficients (MFCC) and Linear Predictive Coding (LPC) coefficients expressing the basic speech features are developed. Combined use of cepstrals of MFCC and LPC in speech recognition system is suggested to improve the reliability of speech recognition system. To this end, the recognition system is divided into MFCC and LPC-based recognition subsystems. The training and recognition processes are realized in both subsystems separately, and recognition system gets the decision being the same results of each subsystems. This results in decrease of error rate during recognition. The training and recognition processes are realized by artificial neural networks in the automatic speech recognition system. The neural networks are trained by the conjugate gradient method. In the paper the problems observed by the number of speech features at training the neural networks of MFCC and LPC-based speech recognition subsystems are investigated. The variety of results of neural networks trained from different initial points in training process is analyzed. Methodology of combined use of neural networks trained from different initial points in speech recognition system is suggested to improve the reliability of recognition system and increase the recognition quality, and obtained practical results are shown.

Keywords: Speech recognition, cepstral analysis, Voice activation detection algorithm, Mel Frequency Cepstral Coefficients, features of speech, Cepstral Mean Subtraction, neural networks, Linear Predictive Coding.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 858

Authors: S. Firouzi, M.R. Alizadeh, S. Minaei

Abstract:

A study was carried out at the Rice Research Institute of Iran (RRII) to investigate the effect of rollers differential peripheral speed of commercial rubber roll husker and paddy moisture content on the husking index and percentage of broken rice. The experiment was conducted at six levels of rollers differential speed (1.5, 2.2, 2.9, 3.6, 4.3 and 5 m/s) and three levels of paddy moisture content (8-9, 10-11 and 12-13% w.b.). Two common paddy varieties namely, Binam and Khazer, were selected for this study. Results revealed that the effect of rollers differential speed and moisture content significantly (P<0.01) affected percentage of broken brown rice and paddy husking index. Average broken kernel percentage increased from 13 to 14.61% while husking index decreased from 71.64 to 61.81%, as paddy moisture content increased from 8-9 to 12-13%. It was observed that amount of broken rice decreased from 18.83 to 9.97%, when rollers differential speed varied from 1.5 to 5 m/s, while the husking index initially increased and then started to decrease. The mean value of husking index for Khazar variety (64.71%) was significantly lower than that for Binam variety (69.2%). It was concluded that rollers differential speed of 2.9 m/s and moisture content of 8-9% was the most appropriate combination for paddy husking of Binam and Khazar varieties in rubber roll husker.

Keywords: husking index, moisture content, paddy, rubber roll husker.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 3231

Authors: M. Israil Hossain, T. P. Tiwari, Ashrafuzzaman Gulandaz, Nusrat Jahan

Abstract:

Two-wheel tractor (power tiller) is a common tillage tool in Bangladesh agriculture for easy access in fragmented land with affordable price of small farmers. Traditional maize sheller needs to be carried from place to place by hooking with two-wheel tractor (2WT) and set up again for shelling operation which takes longer time for preparation of maize shelling. The mobile maize sheller eliminates the transportation problem and can start shelling operation instantly any place as it is attached together with 2WT. It is counterclockwise rotating cylinder, axial flow type sheller, and grain separated with a frictional force between spike tooth and concave. The maize sheller is attached with nuts and bolts in front of the engine base of 2WT. The operating power of the sheller comes from the fly wheel of the engine of the tractor through ‘V” belt pulley arrangement. The average shelling capacity of the mobile sheller is 2.0 t/hr, broken kernel 2.2%, and shelling efficiency 97%. The average maize shelling cost is Tk. 0.22/kg and traditional custom hire rate is Tk.1.0/kg, respectively (1 US$=Tk.78.0). The service provider of the 2WT can transport the mobile maize sheller long distance in operator’s seating position. The manufacturers started the fabrication of mobile maize sheller. This mobile maize sheller is also compatible for the other countries where 2WT is available for farming operation.

Keywords: Cost effective, mobile maize sheller, maize shelling capacity, small farmers, two-wheel tractor.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 857

Authors: M. Javanmard

Abstract:

This study was aimed at analyzing the effects of packaging (MAP) and preservation conditions on the packaged fresh walnut kernel quality. The central composite plan was used for evaluating the effect of oxygen (0–10%), carbon dioxide (0-10%), and temperature (4-26 °C) on qualitative characteristics of walnut kernels. Also, the response level technique was used to find the optimal conditions for interactive effects of factors, as well as estimating the best conditions of process using least amount of testing. Measured qualitative parameters were: peroxide index, color, decreased weight, mould and yeast counting test, and sensory evaluation. The results showed that the defined model for peroxide index, color, weight loss, and sensory evaluation is significant (p < 0.001), so that increase of temperature causes the peroxide value, color variation, and weight loss to increase and it reduces the overall acceptability of walnut kernels. An increase in oxygen percentage caused the color variation level and peroxide value to increase and resulted in lower overall acceptability of the walnuts. An increase in CO₂ percentage caused the peroxide value to decrease, but did not significantly affect other indices (p ≥ 0.05). Mould and yeast were not found in any samples. Optimal packaging conditions to achieve maximum quality of walnuts include: 1.46% oxygen, 10% carbon dioxide, and temperature of 4 °C.

Keywords: Shelled walnut, MAP, quality, storage temperature.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1074

Authors: Guarry Montrose, Ted Soubdhan

Abstract:

In the fight against climate change, the energy consuming building sector must also be taken into account to solve this problem. In this case thermal insulation of buildings using bio-based materials is an interesting solution. Therefore, the thermal performance of some materials of this type has been studied. The advantages of these natural materials of plant origin are multiple, biodegradable, low economic cost, renewable and readily available. The use of biobased materials is widespread in the building sector in order to replace conventional insulation materials with natural materials. Vegetable fibers are very important because they have good thermal behaviour and good insulating properties. The aim of using bio-sourced materials is in line with the logic of energy control and environmental protection, the approach is to make the inhabitants of the houses comfortable and reduce their energy consumption (energy efficiency). In this research we will present the results of studies carried out on the thermal conductivity of banana leaves, latan leaves, vetivers fibers, palm kernel fibers, sargassum, coconut leaves, sawdust and bulk sugarcane leaves. The study on thermal conductivity was carried out in two ways, on the one hand using the flash method, and on the other hand a so-called hot box experiment was carried out. We will discuss and highlight a number of influential factors such as moisture and air pockets present in the samples on the thermophysical properties of these materials, in particular thermal conductivity. Finally, the result of a thermal performance test of banana leaves on a roof in Haiti will also be presented in this work.

Keywords: Buildings, insulating properties, natural materials of plant origin, thermal performance.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 862

Authors: Er, A. C., Rosli, M. H., Asmahani A., Mohamad Naim M. R., Harsuzilawati M.

Abstract:

Dengue is a mosquito-borne infection that has peaked to an alarming rate in recent decades. It can be found in tropical and sub-tropical climate. In Malaysia, dengue has been declared as one of the national health threat to the public. This study aimed to map the spatial distributions of dengue cases in the district of Hulu Langat, Selangor via a combination of Geographic Information System (GIS) and spatial statistic tools. Data related to dengue was gathered from the various government health agencies. The location of dengue cases was geocoded using a handheld GPS Juno SB Trimble. A total of 197 dengue cases occurring in 2003 were used in this study. Those data then was aggregated into sub-district level and then converted into GIS format. The study also used population or demographic data as well as the boundary of Hulu Langat. To assess the spatial distribution of dengue cases three spatial statistics method (Moran-s I, average nearest neighborhood (ANN) and kernel density estimation) were applied together with spatial analysis in the GIS environment. Those three indices were used to analyze the spatial distribution and average distance of dengue incidence and to locate the hot spot of dengue cases. The results indicated that the dengue cases was clustered (p < 0.01) when analyze using Moran-s I with z scores 5.03. The results from ANN analysis showed that the average nearest neighbor ratio is less than 1 which is 0.518755 (p < 0.0001). From this result, we can expect the dengue cases pattern in Hulu Langat district is exhibiting a cluster pattern. The z-score for dengue incidence within the district is -13.0525 (p < 0.0001). It was also found that the significant spatial autocorrelation of dengue incidences occurs at an average distance of 380.81 meters (p < 0.0001). Several locations especially residential area also had been identified as the hot spots of dengue cases in the district.

Keywords: Dengue, geographic information system (GIS), spatial analysis, spatial statistics

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 5286

Authors: Elia Favarelli, Enrico Testi, Andrea Giorgetti

Abstract:

Autonomous structural health monitoring (SHM) of many structures and bridges became a topic of paramount importance for maintenance purposes and safety reasons. This paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data and compare different feature extraction schemes to increase the accuracy and reduce the amount of data collected. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric data in both standard and damaged conditions. The proposed framework starts from the first four fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by time-domain filtering (tracking). The fundamental frequencies extracted are then fed to a dimensionality reduction block implemented through two different approaches: feature selection (intelligent multiplexer) that tries to estimate the most reliable frequencies based on the evaluation of some statistical features (i.e., entropy, variance, kurtosis), and feature extraction (auto-associative neural network (ANN)) that combine the fundamental frequencies to extract new damage sensitive features in a low dimensional feature space. Finally, one-class classification (OCC) algorithms perform anomaly detection, trained with standard condition points, and tested with normal and anomaly ones. In particular, principal component analysis (PCA), kernel principal component analysis (KPCA), and autoassociative neural network (ANN) are presented and their performance are compared. It is also shown that, by evaluating the correct features, the anomaly can be detected with accuracy and an F1 score greater than 95%.

Keywords: Anomaly detection, dimensionality reduction, frequencies selection, modal analysis, neural network, structural health monitoring, vibration measurement.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 633

Authors: Isam A. H. Al Zubaidi

Abstract:

Low sulfur diesel like fuel oil was produced from a quick remediation process of waste oil sludge (WOS). This quick process will reduce the volume of the WOS in petroleum refineries as well as oil fields by transferring the waste to more beneficial product. The practice includes mixing process of WOS with commercial diesel fuel. Different ratios of WOS to diesel fuel were prepared ranging 1:1 to 20:1 by mass. The mixture was continuously mixed for 10 minutes using a bench-type overhead stirrer, and followed by the filtration process to separate the soil waste from filtrate oil product. The quantity and the physical properties of the oil filtrate were measured. It was found that the addition of up to 15% WOS to diesel fuel was accepted without dramatic changes to the properties of diesel fuel. The amount of WOS was decreased by about 60% by mass. This means that about 60% of the mass of sludge was recovered as light fuel oil. The physical properties of the resulting fuel from 10% sludge mixing ratio showed that the specific gravity, ash content, carbon residue, asphaltene content, viscosity, diesel index, cetane number, and calorific value were affected slightly. The color was changed to light black. The sulfur content was increased also. This requires another process to reduce the sulfur content of resulting light fuel. A desulfurization process was achieved using adsorption techniques with activated biomaterial to reduce the sulfur content to acceptable limits. Adsorption process by ZnCl2 activated date palm kernel powder was effective for improvement of the physical properties of diesel like fuel. The final sulfur content was increased to 0.185 wt%. This diesel like fuel can be used in all tractors, buses, tracks inside and outside the refineries. The solid remaining seems to be smooth and can be mixed with asphalt mixture for asphalting the roads or can be used with other materials as asphalt coating material for constructed buildings. Through this process, valuable fuel has been recovered, and the amount of waste material had decreased.

Keywords: Oil sludge, diesel fuel, blending process, filtration process.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 250

Authors: Md. Z. Alam, Amal A. Elgharbawy, Muhammad Moniruzzaman, Nassereldeen A. Kabbashi, Parveen Jamal

Abstract:

The enzymatic hydrolysis of lignocellulosic biomass is one of the obstacles in the process of sugar production, due to the presence of lignin that protects the cellulose molecules against cellulases. Although the pretreatment of lignocellulose in ionic liquid (IL) system has been receiving a lot of interest; however, it requires IL removal with an anti-solvent in order to proceed with the enzymatic hydrolysis. At this point, introducing a compatible cellulase enzyme seems more efficient in this process. A cellulase enzyme that was produced by Trichoderma reesei on palm kernel cake (PKC) exhibited a promising stability in several ILs. The enzyme called PKC-Cel was tested for its optimum pH and temperature as well as its molecular weight. One among evaluated ILs, 1,3-diethylimidazolium dimethyl phosphate [DEMIM] DMP was applied in this study. Evaluation of six factors was executed in Stat-Ease Design Expert V.9, definitive screening design, which are IL/ buffer ratio, temperature, hydrolysis retention time, biomass loading, cellulase loading and empty fruit bunches (EFB) particle size. According to the obtained data, IL-enzyme system shows the highest sugar concentration at 70 °C, 27 hours, 10% IL-buffer, 35% biomass loading, 60 Units/g cellulase and 200 μm particle size. As concluded from the obtained data, not only the PKC-Cel was stable in the presence of the IL, also it was actually stable at a higher temperature than its optimum one. The reducing sugar obtained was 53.468±4.58 g/L which was equivalent to 0.3055 g reducing sugar/g EFB. This approach opens an insight for more studies in order to understand the actual effect of ILs on cellulases and their interactions in the aqueous system. It could also benefit in an efficient production of bioethanol from lignocellulosic biomass.

Keywords: Cellulase, hydrolysis, lignocellulose, pretreatment, stability.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1434

Authors: Benoît Mary, Thibaut Gras, Gaëtan Fagot, Yvon Goth, Ilyes Mnassri-Cetim

Abstract:

A Gerotor pump is composed of an external and internal gear with conjugate cycloidal profiles. From suction to delivery ports, the fluid is transported inside cavities formed by teeth and driven by the shaft. From a geometric and conceptional side it is worth to note that the internal gear has one tooth less than the external one. Simcenter Amesim v.16 includes a new submodel for modelling the hydraulic Gerotor pumps behavior (THCDGP0). This submodel considers leakages between teeth tips using Poiseuille and Couette flows contributions. From the 3D CAD model of the studied pump, the “CAD import” tool takes out the main geometrical characteristics and the submodel THCDGP0 computes the evolution of each cavity volume and their relative position according to the suction or delivery areas. This module, based on international publications, presents robust results up to 6 000 rpm for pressure greater than atmospheric level. For higher rotational speeds or lower pressures, oil aeration and cavitation effects are significant and highly drop the pump’s performance. The liquid used in hydraulic systems always contains some gas, which is dissolved in the liquid at high pressure and tends to be released in a free form (i.e. undissolved as bubbles) when pressure drops. In addition to gas release and dissolution, the liquid itself may vaporize due to cavitation. To model the relative density of the equivalent fluid, modified Henry’s law is applied in Simcenter Amesim v.16 to predict the fraction of undissolved gas or vapor. Three parietal pressure sensors have been set up upstream from the pump to estimate the sound speed in the oil. Analytical models have been compared with the experimental sound speed to estimate the occluded gas content. Simcenter Amesim v.16 model was supplied by these previous analyses marks which have successfully improved the simulations results up to 14 000 rpm. This work provides a sound foundation for designing the next Gerotor pump generation reaching high rotation range more than 25 000 rpm. This improved module results will be compared to tests on this new pump demonstrator.

Keywords: Gerotor pump, high speed, simulations, aeronautic, aeration, cavitation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 502

Authors: Qiuyu Dai, Haochong Zhang, Xiangrong Liu

Abstract:

Efficient matrix-vector multiplication with diagonal sparse matrices is pivotal in a multitude of computational domains, ranging from scientific simulations to machine learning workloads. When encoded in the conventional Diagonal (DIA) format, these matrices often induce computational overheads due to extensive zero-padding and non-linear memory accesses, which can hamper the computational throughput, and elevate the usage of precious compute and memory resources beyond necessity. The ’DIA-Adaptive’ approach, a methodological enhancement introduced in this paper, confronts these challenges head-on by leveraging the advanced parallel instruction sets embedded within Machine Learning Units (MLUs). This research presents a thorough analysis of the DIA-Adaptive scheme’s efficacy in optimizing Sparse Matrix-Vector Multiplication (SpMV) operations. The scope of the evaluation extends to a variety of hardware architectures, examining the repercussions of distinct thread allocation strategies and cluster configurations across multiple storage formats. A dedicated computational kernel, intrinsic to the DIA-Adaptive approach, has been meticulously developed to synchronize with the nuanced performance characteristics of MLUs. Empirical results, derived from rigorous experimentation, reveal that the DIA-Adaptive methodology not only diminishes the performance bottlenecks associated with the DIA format but also exhibits pronounced enhancements in execution speed and resource utilization. The analysis delineates a marked improvement in parallelism, showcasing the DIA-Adaptive scheme’s ability to adeptly manage the interplay between storage formats, hardware capabilities, and algorithmic design. The findings suggest that this approach could set a precedent for accelerating SpMV tasks, thereby contributing significantly to the broader domain of high-performance computing and data-intensive applications.

Keywords: Adaptive method, DIA, diagonal sparse matrices, MLU, sparse matrix-vector multiplication.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 117

Authors: Jaeyoung Lee

Abstract:

Autonomous driving systems require high reliability to provide people with a safe and comfortable driving experience. However, despite the development of a number of vehicle sensors, it is difficult to always provide high perceived performance in driving environments that vary from time to season. The image segmentation method using deep learning, which has recently evolved rapidly, provides high recognition performance in various road environments stably. However, since the system controls a vehicle in real time, a highly complex deep learning network cannot be used due to time and memory constraints. Moreover, efficient networks are optimized for GPU environments, which degrade performance in embedded processor environments equipped simple hardware accelerators. In this paper, a semantic segmentation network, matrix multiplication accelerator network (MMANet), optimized for matrix multiplication accelerator (MMA) on Texas instrument digital signal processors (TI DSP) is proposed to improve the recognition performance of autonomous driving system. The proposed method is designed to maximize the number of layers that can be performed in a limited time to provide reliable driving environment information in real time. First, the number of channels in the activation map is fixed to fit the structure of MMA. By increasing the number of parallel branches, the lack of information caused by fixing the number of channels is resolved. Second, an efficient convolution is selected depending on the size of the activation. Since MMA is a fixed, it may be more efficient for normal convolution than depthwise separable convolution depending on memory access overhead. Thus, a convolution type is decided according to output stride to increase network depth. In addition, memory access time is minimized by processing operations only in L3 cache. Lastly, reliable contexts are extracted using the extended atrous spatial pyramid pooling (ASPP). The suggested method gets stable features from an extended path by increasing the kernel size and accessing consecutive data. In addition, it consists of two ASPPs to obtain high quality contexts using the restored shape without global average pooling paths since the layer uses MMA as a simple adder. To verify the proposed method, an experiment is conducted using perfsim, a timing simulator, and the Cityscapes validation sets. The proposed network can process an image with 640 x 480 resolution for 6.67 ms, so six cameras can be used to identify the surroundings of the vehicle as 20 frame per second (FPS). In addition, it achieves 73.1% mean intersection over union (mIoU) which is the highest recognition rate among embedded networks on the Cityscapes validation set.

Keywords: Edge network, embedded network, MMA, matrix multiplication accelerator and semantic segmentation network.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 401