Search results for: parallel architectures

Authors: Cheikh Bamba Dione, Alla Lo, Elhadji Mamadou Nguer, Siley O. Ba

Abstract:

In this paper, we propose two neural machine translation (NMT) systems (French-to-Wolof and Wolof-to-French) based on sequence-to-sequence with attention and transformer architectures. We trained our models on a parallel French-Wolof corpus of about 83k sentence pairs. Because of the low-resource setting, we experimented with advanced methods for handling data sparsity, including subword segmentation, back translation, and the copied corpus method. We evaluate the models using the BLEU score and find that transformer outperforms the classic seq2seq model in all settings, in addition to being less sensitive to noise. In general, the best scores are achieved when training the models on word-level-based units. For subword-level models, using back translation proves to be slightly beneficial in low-resource (WO) to high-resource (FR) language translation for the transformer (but not for the seq2seq) models. A slight improvement can also be observed when injecting copied monolingual text in the target language. Moreover, combining the copied method data with back translation leads to a substantial improvement of the translation quality.

Keywords: backtranslation, low-resource language, neural machine translation, sequence-to-sequence, transformer, Wolof

Procedia PDF Downloads 113

Authors: A. Abbasi Moshaii, M. Soltan Rezaee, M. Mohammadi Moghaddam

Abstract:

Control of some mechanisms is hard because of their complex dynamic equations. If part of the complexity is resulting from uncertainties, an efficient way for solving that is robust control. By this way, the control procedure could be simple and fast and finally, a simple controller can be designed. One kind of these mechanisms is 3-RRR which is a parallel mechanism and has three revolute joints. This paper aims to robust control a 3-RRR planner mechanism and it presents that this could be used for other mechanisms. So, a significant problem in mechanisms control could be solved. The relevant diagrams are drawn and they show the correctness of control process.

Keywords: 3-RRR, dynamic equations, mechanisms control, structural uncertainty

Procedia PDF Downloads 524

Authors: Andrew Hardman, Alistair T. McIlhagger, Edward Archer

Abstract:

A degree of research has been undertaken in the determination of 3D textile preforms behaviour to compression with direct comparison to 2D counterparts. Multiscale simulations have been developed to try and accurately analyse the behaviour of varying architectures post-consolidation. However, further understanding is required to experimentally identify the mechanisms and deformations that exist upon conforming to a complex contour. Due to the complexity of 3D textile preforms, determination of yarn behaviour to a complex contour is assessed through consolidation by means of vacuum assisted resin transfer moulding (VARTM), and the resulting mechanisms are investigated by micrograph analysis. Varying architectures; with known areal densities, pic density and thicknesses are assessed for a cohesive study. The resulting performance of each is assessed qualitatively as well as quantitatively from the perspective of material in terms of the change in representative unit cell (RVE) across the curved beam contour, in crimp percentage, tow angle, resin rich areas and binder distortion. A novel textile is developed from the resulting analysis to overcome the observed deformations.

Keywords: comformability, compression, binder architecture, 3D weaving, textile preform

Procedia PDF Downloads 139

Authors: Albert Bou, Sebastian Dittert, Gianni de Fabritiis

Abstract:

Advancing reinforcement learning (RL) requires tools that are flexible enough to easily prototype new methods while avoiding impractically slow experimental turnaround times. To match the first requirement, the most popular RL libraries advocate for highly modular agent composability, which facilitates experimentation and development. To solve challenging environments within reasonable time frames, scaling RL to large sampling and computing resources has proved a successful strategy. However, this capability has been so far difficult to combine with modularity. In this work, we explore design choices to allow agent composability both at a local and distributed level of execution. We propose a versatile approach that allows the definition of RL agents at different scales through independent, reusable components. We demonstrate experimentally that our design choices allow us to reproduce classical benchmarks, explore multiple distributed architectures, and solve novel and complex environments while giving full control to the user in the agent definition and training scheme definition. We believe this work can provide useful insights to the next generation of RL libraries.

Keywords: deep reinforcement learning, Python, PyTorch, distributed training, modularity, library

Procedia PDF Downloads 54

Authors: Wen Feiming

Abstract:

As the ecological barrier of the Yangtze River, the ecological security of the Parallel Range-Valley area is very important. However, the unique geomorphic features aggravate the contradiction between man and land, resulting in the encroachment of ecological space. In recent years , relevant researches has focused on the single field of land science, ecology and landscape ecology, and it is difficult to systematically reflect the regularities of distribution and evolution trends of ecological patterns in the process of urban development. Therefore, from the perspective of "Production-Living-Ecological space", using spatial analysis methods such as Remote Sensing (RS) and Geographic Information Systems (GIS), this paper analyzes the evolution characteristics and driving factors of the ecological pattern of mountain towns in the parallel range-valley region from the aspects of land use structure, change rate, transformation relationship, and spatial correlation. It is concluded that the ecological pattern of mountain towns presents a trend from expansion and diffusion to agglomeration, and the dynamic spatial transfer is a trend from artificial transformation to the natural origin, while the driving effect analysis shows the significant characteristics of terrain attraction and construction barrier. Finally, combined with the evolution characteristics and driving mechanism, the evolution modes of "mountain area - concentrated growth", "trough area - diffusion attenuation" and "flat area - concentrated attenuation" are summarized, and the differentiated zoning and stratification ecological planning strategies are proposed here, in order to provide the theoretical basis for the sustainable development of mountain towns in parallel range-valley areas.

Keywords: parallel range-valley, ecological pattern, evolution characteristics, driving factors

Procedia PDF Downloads 64

Authors: Rachid Ziteuni, Selt Omar

Abstract:

In this paper, we propose a new polynomial metaheuristic elaboration (tabu search) for solving scheduling problems. This method allows us to solve the scheduling problem of n tasks on m identical parallel machines with unavailability periods. This problem is NP-complete in the strong sens and finding an optimal solution appears unlikely. Note that all data in this problem are integer and deterministic. The performance criterion to optimize in this problem which we denote Pm/N-c/summs of (wjCj) is the weighted sum of the end dates of tasks.

Keywords: scheduling, parallel identical machines, unavailability periods, metaheuristic, tabu search

Procedia PDF Downloads 298

Authors: Umair Ashraf, Kjetil Uhlen, Sverre Eriksen, Nadeem Jelani

Abstract:

Although advancement in technology has increased the reliability and ease of work in ship power system, but these advancements are also adding complexities. Ever increasing non linear loads, like power electronics (PE) devices effect the stability of the system. Frequent load variations and complex load dynamics are due to the frequency converters and motor drives, these problem are more prominent when system is connected with the weak grid. In the ship power system major consumers are thruster motors for the propulsion. For the control operation of these motors variable frequency drives (VFD) are used, mostly VFDs operate on nominal voltage of the system. Some of the consumers in ship operate on lower voltage than nominal, these consumers got supply through step down transformers. In this paper the vector control scheme is used for the control of both rectifier and inverter, parallel operation of the synchronous generators is also demonstrated. The simulation have been performed with induction motor as load on VFD and parallel RLC load. Fault analysis has been performed first for the system which do not have VFD and then for the system with VFD. Three phase to the ground, single phase to the ground fault were implemented and behavior of the system in both the cases was observed.

Keywords: non-linear load, power electronics, parallel operating generators, pulse width modulation, variable frequency drives, voltage source converters, weak grid

Procedia PDF Downloads 547

Authors: Hasan Al Bashar Abul Ulayee, AKM Saifun Nabi, MD Mesbah-Ul-Awal

Abstract:

Alert system for disaster management is common now a day and can play a vital role reducing devastation and saves lives and costs. An alert in right time can save thousands of human life, help to take shelter, manage other assets including live stocks and above all, a right time alert will help to take preparation to face and early recovery of the situation. In a country like Bangladesh where populations is more than 170 million and always facing different types of natural calamities and disasters, an early right time alert is very effective and implementation of alert system is challenging. The challenge comes from the time constraint of alerting the huge number of population. The other method of existing disaster management pre alert is traditional, sequential and non-selective so efficiency is not good enough. This paper describes a way by which alert can be provided to maximum number of people within the short time bound using parallel processing as well as random selection of selective target audience.

Keywords: alert system, Bangladesh, disaster management, parallel processing, SMS

Procedia PDF Downloads 443

Authors: Sergey Kopysov, Nikita Nedozhogin, Leonid Tonkov

Abstract:

The article presents a parallel iterative solver for large sparse linear systems which can be used on a heterogeneous platform. Traditionally, the problem of solving linear systems does not scale well on multi-CPU/multi-GPUs clusters. For example, most of the attempts to implement the classical conjugate gradient method were at best counted in the same amount of time as the problem was enlarged. The paper proposes the pipelined variant of the conjugate gradient method (PCG), a formulation that is potentially better suited for hybrid CPU/GPU computing since it requires only one synchronization point per one iteration instead of two for standard CG. The standard and pipelined CG methods need the vector entries generated by the current GPU and other GPUs for matrix-vector products. So the communication between GPUs becomes a major performance bottleneck on multi GPU cluster. The article presents an approach to minimize the communications between parallel parts of algorithms. Additionally, computation and communication can be overlapped to reduce the impact of data exchange. Using the pipelined version of the CG method with one synchronization point, the possibility of asynchronous calculations and communications, load balancing between the CPU and GPU for solving the large linear systems allows for scalability. The algorithm is implemented with the combined use of technologies: MPI, OpenMP, and CUDA. We show that almost optimum speed up on 8-CPU/2GPU may be reached (relatively to a one GPU execution). The parallelized solver achieves a speedup of up to 5.49 times on 16 NVIDIA Tesla GPUs, as compared to one GPU.

Keywords: conjugate gradient, GPU, parallel programming, pipelined algorithm

Procedia PDF Downloads 129

Authors: Wojcieh Drożdż, Artur R. Stefankiewicz

Abstract:

Ambidentate ligands can be described as organic structures possessing two different types of coordination units within a single molecule. These features enable the coordination of two different metal ions, which can directly affect the properties of obtained complexes as well as further application. In the current research, we focused on a β-diketone ligand containing terminally located pyridine units in order to assemble cage-like architectures. This will be possible due to the peculiar geometry of the proposed ligands, called "banana-shape", widely used in the synthesis of sophisticated metallosupramolecular architectures. Each of the coordination units plays an important role in cage assembly. Pyridine units enable the coordination of square-planar metal ions (Pd²⁺, Pt²⁺), forming a positively charged cage. On the other hand, the β-diketone group provides the possibility of post-modification, including the introduction of additional functional groups with specific properties (sensing, catalytic, etc.). Such obtained cages are of great interest due to their application potential, including storage or transport of guest molecules, selective detection/separation of analytes as well as efficient catalytic processes.

Keywords: metalloligands, coordination cages, nanoreactors, β-diketonate complexes

Procedia PDF Downloads 43

Authors: Ana Elvira Ojanguren Lopez, Javier Martin Arista

Abstract:

The aims of this paper are to present the automatisation procedure adopted in the implementation of a parallel corpus of Old English, as well as, to assess the progress of automatisation with respect to tagging, annotation, and lemmatisation. The corpus consists of an aligned parallel text with word-for-word comparison Old English-English that provides the Old English segment with inflectional form tagging (gloss, lemma, category, and inflection) and lemma annotation (spelling, meaning, inflectional class, paradigm, word-formation and secondary sources). This parallel corpus is intended to fill a gap in the field of Old English, in which no parallel and/or lemmatised corpora are available, while the average amount of corpus annotation is low. With this background, this presentation has two main parts. The first part, which focuses on tagging and annotation, selects the layouts and fields of lexical databases that are relevant for these tasks. Most information used for the annotation of the corpus can be retrieved from the lexical and morphological database Nerthus and the database of secondary sources Freya. These are the sources of linguistic and metalinguistic information that will be used for the annotation of the lemmas of the corpus, including morphological and semantic aspects as well as the references to the secondary sources that deal with the lemmas in question. Although substantially adapted and re-interpreted, the lemmatised part of these databases draws on the standard dictionaries of Old English, including The Student's Dictionary of Anglo-Saxon, An Anglo-Saxon Dictionary, and A Concise Anglo-Saxon Dictionary. The second part of this paper deals with lemmatisation. It presents the lemmatiser Norna, which has been implemented on Filemaker software. It is based on a concordance and an index to the Dictionary of Old English Corpus, which comprises around three thousand texts and three million words. In its present state, the lemmatiser Norna can assign lemma to around 80% of textual forms on an automatic basis, by searching the index and the concordance for prefixes, stems and inflectional endings. The conclusions of this presentation insist on the limits of the automatisation of dictionary-based annotation in a parallel corpus. While the tagging and annotation are largely automatic even at the present stage, the automatisation of alignment is pending for future research. Lemmatisation and morphological tagging are expected to be fully automatic in the near future, once the database of secondary sources Freya and the lemmatiser Norna have been completed.

Keywords: corpus linguistics, historical linguistics, old English, parallel corpus

Procedia PDF Downloads 173

Authors: Ahmed Machmoum, Malika El Kyal

Abstract:

We consider iterative parallel multi-splitting method for differential algebraic equations. The main feature of the proposed idea is to use the asynchronous form. We prove that the multi-splitting technique can effectively accelerate the convergent performance of the iterative process. The main characteristic of an asynchronous mode is that the local algorithm not have to wait at predetermined messages to become available. We allow some processors to communicate more frequently than others, and we allow the communication delays tobe substantial and unpredictable. Note that synchronous algorithms in the computer science sense are particular cases of our formulation of asynchronous one.

Keywords: computer, multi-splitting methods, asynchronous mode, differential algebraic systems

Procedia PDF Downloads 524

Authors: Alexander Goncharenko, Andrey Denisov, Sergey Alyamkin, Evgeny Terentev

Abstract:

The neural network quantization is highly desired procedure to perform before running neural networks on mobile devices. Quantization without fine-tuning leads to accuracy drop of the model, whereas commonly used training with quantization is done on the full set of the labeled data and therefore is both time- and resource-consuming. Real life applications require simplification and acceleration of quantization procedure that will maintain accuracy of full-precision neural network, especially for modern mobile neural network architectures like Mobilenet-v1, MobileNet-v2 and MNAS. Here we present a method to significantly optimize training with quantization procedure by introducing the trained scale factors for discretization thresholds that are separate for each filter. Using the proposed technique, we quantize the modern mobile architectures of neural networks with the set of train data of only ∼ 10% of the total ImageNet 2012 sample. Such reduction of train dataset size and small number of trainable parameters allow to fine-tune the network for several hours while maintaining the high accuracy of quantized model (accuracy drop was less than 0.5%). Ready-for-use models and code are available in the GitHub repository.

Keywords: distillation, machine learning, neural networks, quantization

Procedia PDF Downloads 291

Authors: Anton Hristozov, Eric Matson, Eric Dietz, Marcus Rogers

Abstract:

Components are heavily used in many software systems, including robotics systems. The growth of sophistication and diversity of new capabilities for robotic systems presents new challenges to their architectures. Their complexity is growing exponentially with the advent of AI, smart sensors, and the complex tasks they have to accomplish. Such complexity requires a more rigorous approach to the creation, use, and interoperability of software components. The issue is exacerbated because robotic systems are becoming more and more reliant on third-party components for certain functions. In order to achieve this kind of interoperability, including dynamic component replacement, we need a way to standardize their interfaces. A formal approach is desperately needed to specify what an interface of a robotic software component should contain. This study performs an analysis of the issue and presents a universal and generic approach to standardizing component interfaces for robotic systems. Our approach is inspired by well-established robotic architectures such as ROS, PX4, and Ardupilot. The study is also applicable to other software systems that share similar characteristics with robotic systems. We consider the use of JSON or Domain Specific Languages (DSL) development with tools such as Antlr and automatic code and configuration file generation for frameworks such as ROS and PX4. A case study with ROS2 is presented as a proof of concept for the proposed methodology.

Keywords: CPS, robots, software architecture, interface, ROS, autopilot

Procedia PDF Downloads 57

Authors: Fatéma Zahra Benchara, Mohamed Youssfi, Omar Bouattane, Hassan Ouajji, Mohamed Ouadi Bensalah

Abstract:

In this paper, we propose a new distributed environment for High Performance Computing (HPC) based on mobile agents. It allows us to perform parallel programs execution as distributed one over a flexible grid constituted by a cooperative mobile agent team works. The distributed program to be performed is encapsulated on team leader agent which deploys its team workers as Agent Virtual Processing Unit (AVPU). Each AVPU is asked to perform its assigned tasks and provides the computational results which make the data and team works tasks management difficult for the team leader agent and that influence the performance computing. In this work we focused on the implementation of the Mobile Provider Agent (MPA) in order to manage the distribution of data and instructions and to ensure a load balancing model. It grants also some interesting mechanisms to manage the others computing challenges thanks to the mobile agents several skills.

Keywords: image processing, distributed environment, mobile agents, parallel and distributed computing

Procedia PDF Downloads 375

Authors: Hui Li, Xueying Zhao, Ke Ma, Yu Cao, Fan Yang, Qingwen Xu, Wenbin Liu

Abstract:

Soil can often link a person or item to a crime scene, which makes it a valuable evidence in forensic casework. Several techniques have been utilized in forensic soil discrimination in previous studies. Because soil contains a vast number of microbiomes, the analyse of soil microbiomes is expected to be a potential way to characterise soil evidence. In this study, we applied massively parallel sequencing (MPS) to soil bacterial profiling on the Ion Torrent Personal Genome Machine (PGM). Soils from different regions were collected repeatedly. V-region 3 and 4 of Bacterial 16S rRNA gene were detected by MPS. Operational taxonomic units (OTU, 97%) were used to analyse soil bacteria. Several bioinformatics methods (PCoA, NMDS, Metastats, LEfse, and Heatmap) were applied in bacterial profiles. Our results demonstrate that MPS can provide a more detailed picture of the soil microbiomes and the composition of soil bacterial components from different region was individualistic. In conclusion, the utility of soil bacterial profiling via MPS of the 16S rRNA gene has potential value in characterising soil evidences and associating them with their place of origin, which can play an important role in forensic science in the future.

Keywords: bacterial profiling, forensic, massively parallel sequencing, soil evidence

Procedia PDF Downloads 531

Authors: Zhen Peng, Baifeng Wu

Abstract:

Modern computing technology enters the era of parallel computing with the trend of sustainable and scalable parallelism. Single Instruction Multiple Data (SIMD) is an important way to go along with the trend. It is able to gather more and more computing ability by increasing the number of processor cores without the need of modifying the program. Meanwhile, in the field of scientific computing and engineering design, many computation intensive applications are facing the challenge of increasingly large amount of data. Data parallel computing will be an important way to further improve the performance of these applications. In this paper, we take the accurate collision detection in building information modeling as an example. We demonstrate a model for constructing a data parallel algorithm. According to the model, a complex object is decomposed into the sets of simple objects; collision detection among complex objects is converted into those among simple objects. The resulting algorithm is a typical SIMD algorithm, and its advantages in parallelism and scalability is unparalleled in respect to the traditional algorithms.

Keywords: data parallelism, collision detection, single instruction multiple data, building information modeling, continuous scalability

Procedia PDF Downloads 257

Authors: Abhishek Bhardwaj, Manish Kumar Bajpai

Abstract:

An image with its content and schema of colors presents an effective mode of information sharing and processing. By changing its color schema different visions and prospect are discovered by the users. This phenomenon of color transfer is being used by Social media and other channel of entertainment. Reinhard et al’s algorithm was the first one to solve this problem of color transfer. In this paper, we make this algorithm efficient by introducing domain parallelism among different processors. We also comment on the factors that affect the speedup of this problem. In the end by analyzing the experimental data we claim to propose a novel and efficient parallel Reinhard’s algorithm.

Keywords: Reinhard et al’s algorithm, color transferring, parallelism, speedup

Procedia PDF Downloads 583

Authors: Aarushi Agarwal, Tara Singh, I.L Singh, Anju Lata Singh, Trayambak Tiwari

Abstract:

Simple discrimination in near absence of attention has been widely observed. Dual-task studies with natural scenes studies have been claimed as being preattentive in nature that facilitated categorization simultaneously with the attentional demanding task. So in this study, multiple images at the periphery are presented, initiating parallel processing in near absence of attention. For the central demanding task rotated letters were presented in both conditions, while in periphery natural and animal images were presented. To understand the breakpoint of ability to perform in near absence of attention one, two and three peripheral images were presented simultaneously with central task and subjects had to respond when all belong to the same category. Individual participant performance did not show a significant difference in both conditions central and peripheral task when the single peripheral image was shown. In case of two images high-level parallel processing could take place with little attentional resources. The eye tracking results supports the evidence as no major saccade was made in a large number of trials. Three image presentations proved to be a breaking point of the capacities to perform outside attentional assistance as participants showed a confused eye gaze pattern which failed to make the natural and animal image discriminations. Thus, we can conclude attention and awareness being independent mechanisms having limited capacities.

Keywords: attention, dual task pardigm, parallel processing, break point, saccade

Procedia PDF Downloads 193

Authors: Nickolay Jelev, Andy Keane, Carren Holden, András Sóbester

Abstract:

In aircraft design, the jump from the conceptual to preliminary design stage introduces a level of complexity which cannot be realistically handled by a single optimiser, be that a human (chief engineer) or an algorithm. The design process is often partitioned along disciplinary lines, with each discipline given a level of autonomy. This introduces a number of challenges including, but not limited to: coupling of design variables; coordinating disciplinary teams; handling of large amounts of analysis data; reaching an acceptable design within time constraints. A number of classical Multidisciplinary Design Optimisation (MDO) architectures exist in academia specifically designed to address these challenges. Their limited use in the industrial aircraft design process has inspired the authors of this paper to develop an alternative strategy based on well established ideas from Decision Support Systems. The proposed rule based architecture sacrifices possibly elusive guarantees of convergence for an attractive return in simplicity. The method is demonstrated on analytical and aircraft design test cases and its performance is compared to a number of classical distributed MDO architectures.

Keywords: Multidisciplinary Design Optimisation, Rule Based Architecture, Aircraft Design, Decision Support System

Procedia PDF Downloads 328

Authors: Lixin Tian, Wei Xue

Abstract:

Many jobs in society go underground, such as mine mining, tunnel construction and subways, which are vital to the development of society. Once accidents occur in these places, the interruption of traditional wired communication is not conducive to the development of rescue work. In order to realize the positioning, early warning and command functions of underground personnel and improve rescue efficiency, it is necessary to develop and design an emergency ground communication system. It is easy to be subjected to narrowband interference when performing conventional underground communication. Spreading communication can be used for this problem. However, general spread spectrum methods such as direct spread communication are inefficient, so it is proposed to use parallel combined spread spectrum (PCSS) communication to improve efficiency. The PCSS communication not only has the anti-interference ability and the good concealment of the traditional spread spectrum system, but also has a relatively high frequency band utilization rate and a strong information transmission capability. So, this technology has been widely used in practice. This paper presents a PCSS communication model-multiple detection parallel combined spread spectrum (MDPCSS) communication system. In this paper, the principle of MDPCSS communication system is described, that is, the sequence at the transmitting end is processed in blocks and cyclically shifted to facilitate multiple detection at the receiving end. The block diagrams of the transmitter and receiver of the MDPCSS communication system are introduced. At the same time, the calculation formula of the system bit error rate (BER) is introduced, and the simulation and analysis of the BER of the system are completed. By comparing with the common parallel PCSS communication, we can draw a conclusion that it is indeed possible to reduce the BER and improve the system performance. Furthermore, the influence of different pseudo-code lengths selected on the system BER is simulated and analyzed, and the conclusion is that the larger the pseudo-code length is, the smaller the system error rate is.

Keywords: cyclic shift, multiple detection, parallel combined spread spectrum, PN code

Procedia PDF Downloads 105

Authors: Nazmul Shahadat, Anthony S. Maida

Abstract:

In recent years, hypercomplex inspired neural networks improved deep CNN architectures due to their ability to share weights across input channels and thus improve cohesiveness of representations within the layers. The work described herein studies the effect of replacing existing layers in an Axial Attention ResNet with their quaternion variants that use cross-channel weight sharing to assess the effect on image classification. We expect the quaternion enhancements to produce improved feature maps with more interlinked representations. We experiment with the stem of the network, the bottleneck layer, and the fully connected backend by replacing them with quaternion versions. These modifications lead to novel architectures which yield improved accuracy performance on the ImageNet300k classification dataset. Our baseline networks for comparison were the original real-valued ResNet, the original quaternion-valued ResNet, and the Axial Attention ResNet. Since improvement was observed regardless of which part of the network was modified, there is a promise that this technique may be generally useful in improving classification accuracy for a large class of networks.

Keywords: axial attention, representational networks, weight sharing, cross-channel correlations, quaternion-enhanced axial attention, deep networks

Procedia PDF Downloads 45

Authors: Apoorva Vinod, Archana Mathur, Snehanshu Saha

Abstract:

Though there exists a plethora of Activation Functions (AFs) used in single and multiple hidden layer Neural Networks (NN), their behavior always raised curiosity, whether used in combination or singly. The popular AFs –Sigmoid, ReLU, and Tanh–have performed prominently well for shallow and deep architectures. Most of the time, AFs are used singly in multi-layered NN, and, to the best of our knowledge, their performance is never studied and analyzed deeply when used in combination. In this manuscript, we experiment with multi-layered NN architecture (both on shallow and deep architectures; Convolutional NN and VGG16) and investigate how well the network responds to using two different AFs (Sigmoid-Tanh, Tanh-ReLU, ReLU-Sigmoid) used alternately against a traditional, single (Sigmoid-Sigmoid, Tanh-Tanh, ReLUReLU) combination. Our results show that using two different AFs, the network achieves better accuracy, substantially lower loss, and faster convergence on 4 computer vision (CV) and 15 Non-CV (NCV) datasets. When using different AFs, not only was the accuracy greater by 6-7%, but we also accomplished convergence twice as fast. We present a case study to investigate the probability of networks suffering vanishing and exploding gradients when using two different AFs. Additionally, we theoretically showed that a composition of two or more AFs satisfies Universal Approximation Theorem (UAT).

Keywords: activation function, universal approximation function, neural networks, convergence

Procedia PDF Downloads 126

Authors: Parisa Golbayani, Dan Wang, Ionut¸ Florescu

Abstract:

In this work we implement machine learning techniques to financial statement reports in order to asses company’s credit rating. Specifically, the work analyzes the performance of four neural network architectures (MLP, CNN, CNN2D, LSTM) in predicting corporate credit rating as issued by Standard and Poor’s. The paper focuses on companies from the energy, financial, and healthcare sectors in the US. The goal of this analysis is to improve application of machine learning algorithms to credit assessment. To accomplish this, the study investigates three questions. First, we investigate if the algorithms perform better when using a selected subset of important features or whether better performance is obtained by allowing the algorithms to select features themselves. Second, we address the temporal aspect inherent in financial data and study whether it is important for the results obtained by a machine learning algorithm. Third, we aim to answer if one of the four particular neural network architectures considered consistently outperforms the others, and if so under which conditions. This work frames the problem as several case studies to answer these questions and analyze the results using ANOVA and multiple comparison testing procedures.

Keywords: convolutional neural network, long short term memory, multilayer perceptron, credit rating

Procedia PDF Downloads 207

Authors: Mais Haj Qasem, Maen M. Al Assaf, Ali Rodan

Abstract:

Parallel hybrid storage systems consist of a hierarchy of different storage devices that vary in terms of data reading speed performance. As we ascend in the hierarchy, data reading speed becomes faster. Thus, migrating the application’ important data that will be accessed in the near future to the uppermost level will reduce the application I/O waiting time; hence, reducing its execution elapsed time. In this research, we implement trace-driven two-levels parallel hybrid storage system prototype that consists of HDDs and SSDs. The prototype uses data mining techniques to classify application’ data in order to determine its near future data accesses in parallel with the its on-demand request. The important data (i.e. the data that the application will access in the near future) are continuously migrated to the uppermost level of the hierarchy. Our simulation results show that our data migration approach integrated with data mining techniques reduces the application execution elapsed time when using variety of traces in at least to 22%.

Keywords: hybrid storage system, data mining, recurrent neural network, support vector machine

Procedia PDF Downloads 276

Authors: Surupa Shaw, Debjyoti Banerjee

Abstract:

Enhancing heat transfer in compact volumes is a challenge when constrained by cost issues, especially those associated with requirements for minimizing pumping power consumption. This is particularly acute for electronic chip cooling applications. Technological advancements in microelectronics have led to development of chip architectures that involve increased power consumption. As a consequence packaging, technologies are saddled with needs for higher rates of power dissipation in smaller form factors. The increasing circuit density, higher heat flux values for dissipation and the significant decrease in the size of the electronic devices are posing thermal management challenges that need to be addressed with a better design of the cooling system. Maximizing surface area for heat exchanging surfaces (e.g., extended surfaces or “fins”) can enable dissipation of higher levels of heat flux. Fractal structures have been shown to maximize surface area in compact volumes. Self-replicating structures at multiple length scales are called “Fractals” (i.e., objects with fractional dimensions; unlike regular geometric objects, such as spheres or cubes whose volumes and surface area values scale as integer values of the length scale dimensions). Fractal structures are expected to provide an appropriate technology solution to meet these challenges for enhanced heat transfer in the microelectronic devices by maximizing surface area available for heat exchanging fluids within compact volumes. In this study, the effect of different fractal micro-channel architectures and flow structures on the enhancement of transport phenomena in heat exchangers is explored by parametric variation of fractal dimension. This study proposes a model that would enable cost-effective solutions for thermal-fluid transport for energy applications. The objective of this study is to ascertain the sensitivity of various parameters (such as heat flux and pressure gradient as well as pumping power) to variation in fractal dimension. The role of the fractal parameters will be instrumental in establishing the most effective design for the optimum cooling of microelectronic devices. This can help establish the requirement of minimal pumping power for enhancement of heat transfer during cooling. Results obtained in this study show that the proposed models for fractal architectures of microchannels significantly enhanced heat transfer due to augmentation of surface area in the branching networks of varying length-scales.

Keywords: fractals, microelectronics, constructal theory, heat transfer enhancement, pumping power enhancement

Procedia PDF Downloads 290

Authors: Hafiz Gulfam Ahmad, Chuangdong Li, Zeeshan Ahmad

Abstract:

In this paper, we proposed a parallel IDS and honeypot based approach to detect and analyze the unknown and known attack taxonomy for improving the IDS performance and protecting the network from intruders. The main theme of our approach is to record and analyze the intruder activities by using both the low and high interaction honeypots. Our architecture aims to achieve the required goals by combing signature based IDS, honeypots and generate the new signatures. The paper describes the basic component, design and implementation of this approach and also demonstrates the effectiveness of this approach reducing the probability of network attacks.

Keywords: network security, intrusion detection, honeypot, snort, nmap

Procedia PDF Downloads 526

Authors: Yi-cheng Chen, Ming-jum Guo, Yang-ru Chen

Abstract:

Background and Objective: For safety and fair during basketball competition, FIBA proposes the definite level of physical functions in wooden surface system (WSS). There are existing various between different systems in indoor-stadium, so the aim of this study want to know how many effects in different WSS, especially for effects of ground reaction force(GRF) when player jumped. Materials and Methods: 12 participants acted counter-movement jump (CMJ) on 7 different surfaces, include 6 WSSs by 3 types rubber shock absorber pad (SAP) on cross or parallel fixed, and 1 rigid ground. GRFs of takeoff and landing had been recorded from an AMTI force platform when all participants acted vertical CMJs by counter-balance design. All data were analyzed using the one-way ANOVA to evaluate whether the test variable differed significantly between surfaces. The significance level was set at α=0.05. Results: There were non-significance in GRF between surfaces when participants taken off. For GRF of landing, we found WSS with cross fixed SAP are harder than parallel fixed. Although there were also non-significance when participant was landing on cross or parallel fixed surfaces, but there have test variable differed significantly between WSS with parallel fixed to rigid ground. In the study, landing to WSS with the hardest SAP, the GRF also have test variable differed significantly to other WSS. Conclusion: Although official basketball competition is in the WSS certificated by FIBA, there are also exist the various in GRF under takeoff or landing, any player must to warm-up before game starting. Especially, there is unsafe situation when play basketball on uncertificated WSS.

Keywords: wooden surface system, counter-movement jump, ground reaction force, shock absorber pad

Procedia PDF Downloads 410

Authors: Djelloul Chafia, Demagh Rafik, Kareche Toufik

Abstract:

Human activities, made without precautions, accelerate the degradation of the soil structure and reduces its resistance. Operations, such as tunnel construction may exercise an influence more or less permanent on the grounds which surrounded them, these structures alter soil it is necessary to predict their impacts by suitable measures. This research is a numerical analysis that deals the risks and effects due to the weakening of the soil after digging two parallel vertical circular tunnels in urban areas, and suggests forecasting techniques based essentially on the organization of underground space. The simulations are performed using the finite-difference code FLAC in a two-dimensional case and with an elasto-plastic behavior of the soil.

Keywords: sol, weakening, degradation, prevention, tunnel

Procedia PDF Downloads 533

Authors: Parisa Khoshvaght, Mehdi Hosseinzadeh

Abstract:

During the past few years, the Residue Number System (RNS) has been receiving considerable interest due to its parallel and fault-tolerant properties. This system is a useful tool for Digital Signal Processing (DSP) since it can support parallel, carry-free, high-speed and low power arithmetic. One of the drawbacks of Residue Number System is the fractional numbers, that is, the corresponding circuit is very hard to realize in conventional CMOS technology. In this paper, we propose a method in which the numbers of transistors are significantly reduced. The related delay is extremely diminished, in the first glance we use this method to solve concerning problem of one decimal functional number some how this proposition can be extended to generalize the idea. Another advantage of this method is the independency on the kind of moduli.

Keywords: computer arithmetic, residue number system, number system, one-Hot, VLSI

Procedia PDF Downloads 470