Search results for: CNN architectures

Authors: Kamalpreet Kaur, Renu Dhir

Abstract:

Climate prediction is an essential component of climate research, which helps evaluate possible effects on economies, communities, and ecosystems. Climate prediction involves short-term weather prediction, seasonal prediction, and long-term climate change prediction. Climate prediction can use the information gathered from satellites, ground-based stations, and ocean buoys, among other sources. The paper's four architectures, such as ResNet50, VGG19, Inception-v3, and Xception, have been combined using an ensemble approach for overall performance and robustness. An ensemble of different models makes a prediction, and the majority vote determines the final prediction. The various architectures such as ResNet50, VGG19, Inception-v3, and Xception efficiently classify the dataset RSI-CB256, which contains satellite images into cloudy and non-cloudy. The generated ensembled S-E model (Sar-ensembled model) provides an accuracy of 99.25%.

Keywords: climate, satellite images, prediction, classification

Procedia PDF Downloads 73

Authors: Giselle Maggie-Fer Castañeda Lozano

Abstract:

The main objective of this research was to explore the role of neural network architectures in simulating behavioral phenomena as a potential explanation for selective associations, specifically related to biological constraints on learning. Biological constraints on learning refer to the limitations observed in conditioning procedures, where learning is expected to occur. The study involved simulations of five different experiments exploring various phenomena and sources of biological constraints in learning. These simulations included the interaction between response and reinforcer, stimulus and reinforcer, specificity of stimulus-reinforcer associations, species differences, neuroanatomical constraints, and learning in uncontrolled conditions. The overall results demonstrated that by manipulating neural network architectures, conditions can be created to model and explain diverse biological constraints frequently reported in comparative psychology literature as learning typicities. Additionally, the simulations offer predictive content worthy of experimental testing in the pursuit of new discoveries regarding the specificity of learning. The implications and limitations of these findings are discussed. Finally, it is suggested that this research could inaugurate a line of inquiry involving the use of neural networks to study biological factors in behavior, fostering the development of more ethical and precise research practices.

Keywords: comparative psychology, connectionism, conditioning, experimental analysis of behavior, neural networks

Procedia PDF Downloads 71

Authors: Sushish Baral, Pratibha Joshi, Yaman Maharjan

Abstract:

Glaucoma is a chronic eye condition that causes vision loss that is irreversible. Early detection and treatment are critical to prevent vision loss because it can be asymptomatic. For the identification of glaucoma, multiple deep learning algorithms are used. Transformer-based architectures, which use the self-attention mechanism to encode long-range dependencies and acquire extremely expressive representations, have recently become popular. Convolutional architectures, on the other hand, lack knowledge of long-range dependencies in the image due to their intrinsic inductive biases. The aforementioned statements inspire this thesis to look at transformer-based solutions and investigate the viability of adopting transformer-based network designs for glaucoma detection. Using retinal fundus images of the optic nerve head to develop a viable algorithm to assess the severity of glaucoma necessitates a large number of well-curated images. Initially, data is generated by augmenting ocular pictures. After that, the ocular images are pre-processed to make them ready for further processing. The system is trained using pre-processed images, and it classifies the input images as normal or glaucoma based on the features retrieved during training. The Vision Transformer (ViT) architecture is well suited to this situation, as it allows the self-attention mechanism to utilise structural modeling. Extensive experiments are run on the common dataset, and the results are thoroughly validated and visualized.

Keywords: glaucoma, vision transformer, convolutional architectures, retinal fundus images, self-attention, deep learning

Procedia PDF Downloads 191

Authors: Tianyu Gao

Abstract:

Digital architectures use computer-aided design, programming, simulation, and imaging to create virtual forms and physical structures. Today's customers want to know more about their buildings. They want an automatic thermostat to learn their behavior and contact them, such as the doors and windows they want to open with a mobile app. Therefore, the architectural display form is more closely related to the customer's experience. Based on the purpose of building informationization, this paper studies the steel structure design based on BIM. Taking the Zigan office building in Hangzhou as an example, it is divided into four parts, namely, the digital design modulus of the steel structure, the node analysis of the steel structure, the digital production and construction of the steel structure. Through the application of BIM software, the architectural design can be synergized, and the building components can be informationized. Not only can the architectural design be feedback in the early stage, but also the stability of the construction can be guaranteed. In this way, the monitoring of the entire life cycle of the building and the meeting of customer needs can be realized.

Keywords: digital architectures, BIM, steel structure, architectural design

Procedia PDF Downloads 195

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: Jyothirmayee Pola

Abstract:

Ideally, a test-driven development (TDD) or agile or any other process should be able to define and implement performance, scalability, and reliability (PSR) of the product with a higher quality of service (QOS) and should have the ability to fix any PSR issues with lesser cost before it hits the production. Most PSR test strategies for new product introduction (NPI) include assumptions about production load requirements but never accurate. NPE (New product Enhancement) include assumptions for new features that are being developed whilst workload distribution for older features can be derived by analyzing production transactions. This paper talks about how to shift left PSR towards design phase of release management process to get better QOS w.r.t PSR for any product under development. It also explains the ROI for future customer onboarding both for Service Oriented Architectures (SOA) and Microservices architectures and how to define PSR requirements.

Keywords: component PSR, performance engineering, performance tuning, reliability, return on investment, scalability, system PSR

Procedia PDF Downloads 75

Authors: Ousmane Youme, Jean Marie Dembele, Eugene Ezin, Christophe Cambier

Abstract:

In recent years, the convolutional neural networks (CNN) architectures designed by evolution algorithms have proven to be competitive with handcrafted architectures designed by experts. However, these algorithms need a lot of computational power, which is beyond the capabilities of most researchers and engineers. To overcome this problem, we propose an evolution architecture under length constraints. It consists of two algorithms: a search length strategy to find an optimal space and a search architecture strategy based on a genetic algorithm to find the best individual in the optimal space. Our algorithms drastically reduce resource costs and also keep good performance. On the Cifar-10 dataset, our framework presents outstanding performance with an error rate of 5.12% and only 4.6 GPU a day to converge to the optimal individual -22 GPU a day less than the lowest cost automatic evolutionary algorithm in the peer competition.

Keywords: CNN architecture, genetic algorithm, evolution algorithm, length constraints

Procedia PDF Downloads 128

Authors: Brenden M. Shutt, Lubjana Beshaj, Paul L. Goethals, Ambrose Kam

Abstract:

Communication security is of particular interest to military data networks. A relatively novel approach to network security is blockchain, a cryptographically secured distribution ledger with a decentralized consensus mechanism for data transaction processing. Recent advances in blockchain technology have proposed new techniques for both data validation and trust management, as well as different frameworks for managing dataflow. The purpose of this work is to test the feasibility of different blockchain architectures as applied to military command and control networks. Various architectures are tested through discrete-event simulation and the feasibility is determined based upon a blockchain design’s ability to maintain long-term stable performance at industry standards of throughput, network latency, and security. This work proposes a consortium blockchain architecture with a computationally inexpensive consensus mechanism, one that leverages a Proof-of-Identity (PoI) concept and a reputation management mechanism.

Keywords: blockchain, consensus mechanism, discrete-event simulation, fog computing

Procedia PDF Downloads 138

Authors: Sarita Agarwal, Deepika Delsa Dean

Abstract:

Like the task of transferring photo images to artistic images, image-to-image translation aims to translate the data to the imitated data which belongs to the target domain. Neural Style Transfer and CycleGAN are two well-known deep learning architectures used for photo image-to-art image transfer. However, studies involving these two models concentrate on one-to-one domain translation, not one-to-multi domains translation. Our study tries to investigate deep learning architectures, which can be controlled to yield multiple artistic style translation only by adding a conditional vector. We have expanded CycleGAN and constructed Conditional CycleGAN for 5 kinds of categories translation. Our study found that the architecture inserting conditional vector into the middle layer of the Generator could output multiple artistic images.

Keywords: genetic counseling, FMR1 gene, fragile x-associated primary ovarian insufficiency, premutation

Procedia PDF Downloads 130

Authors: Subodh Chandra Shakya, Badal Shrestha, Suni Thapa, Ashutosh Chauhan, Saugat Adhikari

Abstract:

To tackle our problem of clothes identification, we used different architectures of Convolutional Neural Networks. Among different architectures, the outcome from Inception ResNet V2 and MobileNet V2 seemed promising. On comparison of the metrices, we observed that the Inception ResNet V2 slightly outperforms MobileNet V2 for this purpose. So this paper of ours proposes the cloth identifier using Inception ResNet V2 and also contains the comparison between the outcome of ResNet V2 and MobileNet V2. The document here contains the results and findings of the research that we performed on the DeepFashion Dataset. To improve the dataset, we used different image preprocessing techniques like image shearing, image rotation, and denoising. The whole experiment was conducted with the intention of testing the efficiency of convolutional neural networks on cloth identification so that we could develop a reliable system that is good enough in identifying the clothes worn by the users. The whole system can be integrated with some kind of recommendation system.

Keywords: inception ResNet, convolutional neural net, deep learning, confusion matrix, data augmentation, data preprocessing

Procedia PDF Downloads 187

Authors: Rossella Arcucci, Luisa D'Amore, Simone Celestino, Giuseppe Scotti, Giuliano Laccetti

Abstract:

This work is the first dowel in a rather wide research activity in collaboration with Euro Mediterranean Center for Climate Changes, aimed at introducing scalable approaches in Ocean Circulation Models. We discuss designing and implementation of a parallel algorithm for solving the Variational Data Assimilation (DA) problem on Graphics Processing Units (GPUs). The algorithm is based on the fully scalable 3DVar DA model, previously proposed by the authors, which uses a Domain Decomposition approach (we refer to this model as the DD-DA model). We proceed with an incremental porting process consisting of 3 distinct stages: requirements and source code analysis, incremental development of CUDA kernels, testing and optimization. Experiments confirm the theoretic performance analysis based on the so-called scale up factor demonstrating that the DD-DA model can be suitably mapped on GPU architectures.

Keywords: data assimilation, GPU architectures, ocean models, parallel algorithm

Procedia PDF Downloads 412

Authors: Zaki Ahmad Khan, Jamshed Siddiqui, Abdus Samad

Abstract:

The Scheduling and mapping of tasks on a set of processors is considered as a critical problem in parallel and distributed computing system. This paper deals with the problem of dynamic scheduling on a special type of multiprocessor architecture known as Linear Crossed Cube (LCQ) network. This proposed multiprocessor is a hybrid network which combines the features of both linear type of architectures as well as cube based architectures. Two standard dynamic scheduling schemes namely Minimum Distance Scheduling (MDS) and Two Round Scheduling (TRS) schemes are implemented on the LCQ network. Parallel tasks are mapped and the imbalance of load is evaluated on different set of processors in LCQ network. The simulations results are evaluated and effort is made by means of through analysis of the results to obtain the best solution for the given network in term of load imbalance left and execution time. The other performance matrices like speedup and efficiency are also evaluated with the given dynamic algorithms.

Keywords: dynamic algorithm, load imbalance, mapping, task scheduling

Procedia PDF Downloads 449

Authors: Bandhan Dey, Muhsina Bintoon Yiasha, Gulam Sulaman Choudhury

Abstract:

Chest disease is one of the most problematic ailments in our regular life. There are many known chest diseases out there. Diagnosing them correctly plays a vital role in the process of treatment. There are many methods available explicitly developed for different chest diseases. But the most common approach for diagnosing these diseases is through X-ray. In this paper, we proposed a multi-classification deep learning model for diagnosing COVID-19, lung cancer, pneumonia, tuberculosis, and atelectasis from chest X-rays. In the present work, we used the transfer learning method for better accuracy and fast training phase. The performance of three architectures is considered: InceptionV3, VGG-16, and VGG-19. We evaluated these deep learning architectures using public digital chest x-ray datasets with six classes (i.e., COVID-19, lung cancer, pneumonia, tuberculosis, atelectasis, and normal). The experiments are conducted on six-classification, and we found that VGG16 outperforms other proposed models with an accuracy of 95%.

Keywords: deep learning, image classification, X-ray images, Tensorflow, Keras, chest diseases, convolutional neural networks, multi-classification

Procedia PDF Downloads 92

Authors: Tushar K. Routh

Abstract:

If inputs are engineered in certain manners, they can influence deep neural networks’ (DNN) performances by facilitating misclassifications, a phenomenon well-known as adversarial attacks that question networks’ vulnerability. Recent studies have unfolded the relationship between vulnerability of such networks with their complexity. In this paper, the distinctive influence of additional convolutional layers at the decision boundaries of several DNN architectures was investigated. Here, to engineer inputs from widely known image datasets like MNIST, Fashion MNIST, and Cifar 10, we have exercised One Step Spectral Attack (OSSA) and Fast Gradient Method (FGM) techniques. The aftermaths of adding layers to the robustness of the architectures have been analyzed. For reasoning, separation width from linear class partitions and local geometry (curvature) near the decision boundary have been examined. The result reveals that model complexity has significant roles in adjusting relative distances from margins, as well as the local features of decision boundaries, which impact robustness.

Keywords: DNN robustness, decision boundary, local curvature, network complexity

Procedia PDF Downloads 75

Authors: Calvin Ralph, Edward Archer, Alistair McIlhagger

Abstract:

3D woven textile composites continue to emerge as an advanced material for structural applications and composite manufacture due to their bespoke nature, through thickness reinforcement and near net shape capabilities. When 3D woven preforms are produced, they are in their optimal physical state. As 3D weaving is a dry preforming technology it relies on compression of the preform to achieve the desired composite thickness, fibre volume fraction (Vf) and consolidation. This compression of the preform during manufacture results in changes to its thickness and architecture which can often lead to under-performance or changes of the 3D woven composite. Unlike traditional 2D fabrics, the bespoke nature and variability of 3D woven architectures makes it difficult to know exactly how each 3D preform will behave during processing. Therefore, the focus of this study is to investigate the effect of compression on differing 3D woven architectures in terms of structure, crimp or fibre waviness and thickness as well as analysing the accuracy of available software to predict how 3D woven preforms behave under compression. To achieve this, 3D preforms are modelled and compression simulated in Wisetex with varying architectures of binder style, pick density, thickness and tow size. These architectures have then been woven with samples dry compression tested to determine the compressibility of the preforms under various pressures. Additional preform samples were manufactured using Resin Transfer Moulding (RTM) with varying compressive force. Composite samples were cross sectioned, polished and analysed using microscopy to investigate changes in architecture and crimp. Data from dry fabric compression and composite samples were then compared alongside the Wisetex models to determine accuracy of the prediction and identify architecture parameters that can affect the preform compressibility and stability. Results indicate that binder style/pick density, tow size and thickness have a significant effect on compressibility of 3D woven preforms with lower pick density allowing for greater compression and distortion of the architecture. It was further highlighted that binder style combined with pressure had a significant effect on changes to preform architecture where orthogonal binders experienced highest level of deformation, but highest overall stability, with compression while layer to layer indicated a reduction in fibre crimp of the binder. In general, simulations showed a relative comparison to experimental results; however, deviation is evident due to assumptions present within the modelled results.

Keywords: 3D woven composites, compression, preforms, textile composites

Procedia PDF Downloads 135

Authors: Namugenyi Ephrance Eunice, Julianne Sansa Otim, Marco Zennaro, Stephen D. Wolthusen

Abstract:

The Internet of Things (IoT) is a network of connected data processing devices, mechanical and digital machinery, items, animals, or people that may send data across a network without requiring human-to-human or human-to-computer interaction. Each component has sensors that can pick up on specific phenomena, as well as processing software and other technologies that can link to and communicate with other systems and/or devices over the Internet or other communication networks and exchange data with them. IoT is increasingly being used in fields other than consumer electronics, such as public safety, emergency response, industrial automation, autonomous vehicles, the Internet of Medical Things (IoMT), and general environmental monitoring. Consumer-based IoT applications, like smart home gadgets and wearables, are also becoming more prevalent. This paper presents the main IoT deployment areas for environmental monitoring in developing regions and the backhaul options suitable for them. A detailed review of each of the list of papers selected for the study is included in section III of this document. The study includes an overview of existing IoT deployments, the underlying communication architectures, protocols, and technologies that support them. This overview shows that Low Power Wireless Area Networks (LPWANs), as summarized in Table 1, are very well suited for monitoring environment architectures designed for remote locations. LoRa technology, particularly the LoRaWAN protocol, has an advantage over other technologies due to its low power consumption, adaptability, and suitable communication range. The prevailing challenges of the different architectures are discussed and summarized in Table 3 of the IV section, where the main problem is the obstruction of communication paths by buildings, trees, hills, etc.

Keywords: communication technologies, environmental monitoring, Internet of Things, IoT deployment challenges

Procedia PDF Downloads 85

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 166

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 83

Authors: Mariem Saied, Jens Gustedt, Gilles Muller

Abstract:

We present Dido, a source-to-source auto-generation and optimization framework for multi-dimensional stencil computations. It enables a large programmer community to easily and safely implement stencil codes on distributed-memory parallel architectures with Ordered Read-Write Locks (ORWL) as an execution and communication back-end. ORWL provides inter-task synchronization for data-oriented parallel and distributed computations. It has been proven to guarantee equity, liveness, and efficiency for a wide range of applications, particularly for iterative computations. Dido consists mainly of an implicitly parallel domain-specific language (DSL) implemented as a source-level transformer. It captures domain semantics at a high level of abstraction and generates parallel stencil code that leverages all ORWL features. The generated code is well-structured and lends itself to different possible optimizations. In this paper, we enhance Dido to handle both Jacobi and Gauss-Seidel grid traversals. We integrate temporal blocking to the Dido code generator in order to reduce the communication overhead and minimize data transfers. To increase data locality and improve intra-node data reuse, we coupled the code generation technique with the polyhedral parallelizer Pluto. The accuracy and portability of the generated code are guaranteed thanks to a parametrized solution. The combination of ORWL features, the code generation pattern and the suggested optimizations, make of Dido a powerful code generation framework for stencil computations in general, and for distributed-memory architectures in particular. We present a wide range of experiments over a number of stencil benchmarks.

Keywords: stencil computations, ordered read-write locks, domain-specific language, polyhedral model, experiments

Procedia PDF Downloads 127

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 73

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 325

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 92

Authors: Mustafa Elfituri, Jonathan Cook

Abstract:

Recently, graph-based computations have become more important in large-scale scientific computing as they can provide a methodology to model many types of relations between independent objects. They are being actively used in fields as varied as biology, social networks, cybersecurity, and computer networks. At the same time, graph problems have some properties such as irregularity and poor locality that make their performance different than regular applications performance. Therefore, parallelizing graph algorithms is a hard and challenging task. Initial evidence is that standard computer architectures do not perform very well on graph algorithms. Little is known exactly what causes this. The Graph500 benchmark is a representative application for parallel graph-based computations, which have highly irregular data access and are driven more by traversing connected data than by computation. In this paper, we present results from analyzing the performance of various example implementations of Graph500, including a shared memory (OpenMP) version, a distributed (MPI) version, and a hybrid version. We measured and analyzed all the factors that affect its performance in order to identify possible changes that would improve its performance. Results are discussed in relation to what factors contribute to performance degradation.

Keywords: graph computation, graph500 benchmark, parallel architectures, parallel programming, workload characterization.

Procedia PDF Downloads 147

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 353

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 83

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 158

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 235

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 318

Authors: Salim Yusif

Abstract:

Fiber optic technology has revolutionized telecommunications and data transmission, providing unmatched speed, bandwidth, and reliability. This paper presents the latest advancements in fiber optic technology, focusing on innovations in fiber materials, transmission techniques, and network architectures that enhance the performance of high-speed data transmission systems. Key advancements include the development of ultra-low-loss optical fibers, multi-core fibers, advanced modulation formats, and the integration of fiber optics into next-generation network architectures such as Software-Defined Networking (SDN) and Network Function Virtualization (NFV). Additionally, recent developments in fiber optic sensors are discussed, extending the utility of optical fibers beyond data transmission. Through comprehensive analysis and experimental validation, this research offers valuable insights into the future directions of fiber optic technology, highlighting its potential to drive innovation across various industries.

Keywords: fiber optics, high-speed data transmission, ultra-low-loss optical fibers, multi-core fibers, modulation formats, coherent detection, software-defined networking, network function virtualization, fiber optic sensors

Procedia PDF Downloads 61

Authors: Pranav Gunhal

Abstract:

This paper presents an analysis of sentiment on Twitter towards the Karnataka elections held in 2023, utilizing transformer-based models specifically designed for sentiment analysis in Indic languages. Through an innovative data collection approach involving a combination of novel methods of data augmentation, online data preceding the election was analyzed. The study focuses on sentiment classification, effectively distinguishing between positive, negative, and neutral posts while specifically targeting the sentiment regarding the loss of the Bharatiya Janata Party (BJP) or the win of the Indian National Congress (INC). Leveraging high-performing transformer architectures, specifically IndicBERT, coupled with specifically fine-tuned hyperparameters, the AI models employed in this study achieved remarkable accuracy in predicting the INC’s victory in the election. The findings shed new light on the potential of cutting-edge transformer-based models in capturing and analyzing sentiment dynamics within the Indian political landscape. The implications of this research are far-reaching, providing invaluable insights to political parties for informed decision-making and strategic planning in preparation for the forthcoming 2024 Lok Sabha elections in the nation.

Keywords: sentiment analysis, twitter, Karnataka elections, congress, BJP, transformers, Indic languages, AI, novel architectures, IndicBERT, lok sabha elections

Procedia PDF Downloads 84