scholarly journals Optimization of the Convolution Operation to Accelerate Deep Neural Networks in FPGA

2021 ◽  
Vol 35 (6) ◽  
pp. 511-517
Author(s):  
Malathi Devendran ◽  
Indumathi Rajendran ◽  
Vijayakumar Ponnusamy ◽  
Diwakar R. Marur
Keyword(s):  
Neural Networks ◽  
Computation Time ◽  
Input Image ◽  
Optimization Techniques ◽  
Machine Learning Algorithms ◽  
Loop Optimization ◽  
Pixel Intensity ◽  
Loop Unrolling ◽  
Field Programmable ◽  
Four Levels

In recent years, machine learning algorithms related to images have been widely utilized by Convolution Neural Networks (CNN), and it has a high accuracy for recognition of an image. As CNN contains large number of computations, hardware accelerator like Field Programmable Gate Array is employed. Quite 90 % of operations during a CNN involves convolution. The objective of this work is to scale back the computation time to increase the peak, width and the pixel intensity levels in the input image. The execution time of a image processing program is mostly spent on loops. Loop optimization is a process of accelerating speed and reducing the overheads related to loops. It plays a crucial role in improving performance and making effective use of multiprocessing capabilities. Loop unrolling is one of the loop optimization techniques. In our work CNN with four levels of loop unrolling is used. Due to this delay is reduced compared with conventional Xilinix. With the assistance of strides and padding the 40 % of computation time has been reduced and is verified in MATLAB.

scholarly journals Integrating Traditional Machine Learning and Neural Networks for Image Processing

2021 ◽  
Author(s):  
Nikita Laptev ◽  
Vladislav Laptev ◽  
Olga Gerget ◽  
Dmitriy Kolpashchikov
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Image Processing ◽  
Short Term Memory ◽  
Learning Algorithms ◽  
Input Image ◽  
Machine Learning Algorithms ◽  
Neural Network Models ◽  
Fire Source ◽  
Artificial Neural

The article describes a feasibility study to assess the use of neural networks and traditional machine learning algorithms to solve various problems including image processing. A brief description of some algorithms of traditional machine learning, as well as anautomated service for choosing the best method for a specific task, is given. The authors also describe the features of artificial neural networks and the most popular places for theirapplication. An algorithm for solving the problem of detecting fire hazardous objects andlocalizing a fire source in a forest using video sequence frames is presented. The article compares the characteristics of artificial neural network models according to the followingcriteria: underlying architecture, the number of analyzed frames, the size of the input image, the transfer learning model used as a feature vector composing network. Acomparative analysis of traditional machine learning algorithms and neural networks withlong short-term memory in the problem of classification of forest fire hazards is made. A solution to localization of the source of fire based on clustering is described. A hybrid algorithm for finding a fire source in a forest is developed and illustrated.

Segmentasi Citra Mri Menggunakan Deteksi Tepi Untuk Identifikasi Kanker Payudara

2017 ◽  
Vol 15 (2) ◽  
pp. 17 ◽  
Author(s):  
Ervina Varijki ◽  
Bambang Krismono Triwijoyo
Keyword(s):  
Breast Cancer ◽  
Edge Detection ◽  
Object Segmentation ◽  
Computation Time ◽  
Input Image ◽  
Boundary Line ◽  
Cancer Breast ◽  
Average Computation Time ◽  
Breast Cancer Mri ◽  
Image Format

One type of cancer that is capable identified using MRI technology is breast cancer. Breast cancer is still the leading cause of death world. therefore early detection of this disease is needed. In identifying breast cancer, a doctor or radiologist analyzing the results of magnetic resonance image that is stored in the format of the Digital Imaging Communication In Medicine (DICOM). It takes skill and experience sufficient for diagnosis is appropriate, andaccurate, so it is necessary to create a digital image processing applications by utilizing the process of object segmentation and edge detection to assist the physician or radiologist in identifying breast cancer. MRI image segmentation using edge detection to identification of breast cancer using a method stages gryascale change the image format, then the binary image thresholding and edge detection process using the latest Robert operator. Of the20 tested the input image to produce images with the appearance of the boundary line of each region or object that is visible and there are no edges are cut off, with the average computation time less than one minute.

TLBO-FLN: Teaching-Learning Based Optimization of Functional Link Neural Networks for Stock Closing Price Prediction

2020 ◽  
Vol 10 (4) ◽  
pp. 522-532 ◽  
Author(s):  
Sarat Chandra Nayak ◽  
Subhranginee Das ◽  
Mohammad Dilsad Ansari
Keyword(s):  
Neural Networks ◽  
Computational Cost ◽  
Optimization Techniques ◽  
Fine Tuning ◽  
Functional Link ◽  
Price Prediction ◽  
Closing Price ◽  
Teaching Learning Based Optimization ◽  
Artificial Neural ◽  
Teaching Learning

Background and Objective: Stock closing price prediction is enormously complicated. Artificial Neural Networks (ANN) are excellent approximation algorithms applied to this area. Several nature-inspired evolutionary optimization techniques are proposed and used in the literature to search the optimum parameters of ANN based forecasting models. However, most of them need fine-tuning of several control parameters as well as algorithm specific parameters to achieve optimal performance. Improper tuning of such parameters either leads toward additional computational cost or local optima. Methods: Teaching Learning Based Optimization (TLBO) is a newly proposed algorithm which does not necessitate any parameters specific to it. The intrinsic capability of Functional Link Artificial Neural Network (FLANN) to recognize the multifaceted nonlinear relationship present in the historical stock data made it popular and got wide applications in the stock market prediction. This article presents a hybrid model termed as Teaching Learning Based Optimization of Functional Neural Networks (TLBO-FLN) by combining the advantages of both TLBO and FLANN. Results and Conclusion: The model is evaluated by predicting the short, medium, and long-term closing prices of four emerging stock markets. The performance of the TLBO-FLN model is measured through Mean Absolute Percentage of Error (MAPE), Average Relative Variance (ARV), and coefficient of determination (R2); compared with that of few other state-of-the-art models similarly trained and found superior.

scholarly journals Prediction of Sugar Content in Port Wine Vintage Grapes Using Machine Learning and Hyperspectral Imaging

Processes ◽  
2021 ◽  
Vol 9 (7) ◽  
pp. 1241
Author(s):  
Véronique Gomes ◽  
Marco S. Reis ◽  
Francisco Rovira-Más ◽  
Ana Mendes-Ferreira ◽  
Pedro Melo-Pinto
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Hyperspectral Imaging ◽  
Sugar Content ◽  
Hyperspectral Data ◽  
Machine Learning Algorithms ◽  
Port Wine ◽  
Monitoring And Control ◽  
High Quality ◽  
Harvesting Stage

The high quality of Port wine is the result of a sequence of winemaking operations, such as harvesting, maceration, fermentation, extraction and aging. These stages require proper monitoring and control, in order to consistently achieve the desired wine properties. The present work focuses on the harvesting stage, where the sugar content of grapes plays a key role as one of the critical maturity parameters. Our approach makes use of hyperspectral imaging technology to rapidly extract information from wine grape berries; the collected spectra are fed to machine learning algorithms that produce estimates of the sugar level. A consistent predictive capability is important for establishing the harvest date, as well as to select the best grapes to produce specific high-quality wines. We compared four different machine learning methods (including deep learning), assessing their generalization capacity for different vintages and varieties not included in the training process. Ridge regression, partial least squares, neural networks and convolutional neural networks were the methods considered to conduct this comparison. The results show that the estimated models can successfully predict the sugar content from hyperspectral data, with the convolutional neural network outperforming the other methods.

scholarly journals Attention-based deep learning networks for identification of human gait using radar micro-Doppler spectrograms

2021 ◽  
pp. 1-6
Author(s):  
Hannah Garcia Doherty ◽  
Roberto Arnaiz Burgueño ◽  
Roeland P. Trommel ◽  
Vasileios Papanastasiou ◽  
Ronny I. A. Harmanny
Keyword(s):  
Neural Networks ◽  
Feature Vector ◽  
Classification Performance ◽  
Input Image ◽  
Human Gait ◽  
Learning Networks ◽  
Class Label ◽  
Deep Convolutional Neural Networks ◽  
Network Layers ◽  
Feature Dimension

Abstract Identification of human individuals within a group of 39 persons using micro-Doppler (μ-D) features has been investigated. Deep convolutional neural networks with two different training procedures have been used to perform classification. Visualization of the inner network layers revealed the sections of the input image most relevant when determining the class label of the target. A convolutional block attention module is added to provide a weighted feature vector in the channel and feature dimension, highlighting the relevant μ-D feature-filled areas in the image and improving classification performance.

scholarly journals Programmable phase-change metasurfaces on waveguides for multimode photonic convolutional neural network

2021 ◽  
Vol 12 (1) ◽  
Author(s):  
Changming Wu ◽  
Heshan Yu ◽  
Seokhyeong Lee ◽  
Ruoming Peng ◽  
Ichiro Takeuchi ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Phase Change ◽  
Convolutional Neural Network ◽  
Large Scale ◽  
Phase Change Materials ◽  
Refractive Index Change ◽  
Optical Computing ◽  
Machine Learning Algorithms ◽  
Matrix Vector Multiplication

AbstractNeuromorphic photonics has recently emerged as a promising hardware accelerator, with significant potential speed and energy advantages over digital electronics for machine learning algorithms, such as neural networks of various types. Integrated photonic networks are particularly powerful in performing analog computing of matrix-vector multiplication (MVM) as they afford unparalleled speed and bandwidth density for data transmission. Incorporating nonvolatile phase-change materials in integrated photonic devices enables indispensable programming and in-memory computing capabilities for on-chip optical computing. Here, we demonstrate a multimode photonic computing core consisting of an array of programable mode converters based on on-waveguide metasurfaces made of phase-change materials. The programmable converters utilize the refractive index change of the phase-change material Ge2Sb2Te5 during phase transition to control the waveguide spatial modes with a very high precision of up to 64 levels in modal contrast. This contrast is used to represent the matrix elements, with 6-bit resolution and both positive and negative values, to perform MVM computation in neural network algorithms. We demonstrate a prototypical optical convolutional neural network that can perform image processing and recognition tasks with high accuracy. With a broad operation bandwidth and a compact device footprint, the demonstrated multimode photonic core is promising toward large-scale photonic neural networks with ultrahigh computation throughputs.

scholarly journals Real-time classification of hand movements as a basis for intuitive control of grasp neuroprostheses

2020 ◽  
Vol 6 (2) ◽  
Author(s):  
Dmitry Amelin ◽  
Ivan Potapov ◽  
Josep Cardona Audí ◽  
Andreas Kogut ◽  
Rüdiger Rupp ◽  
...  
Keyword(s):  
Neural Networks ◽  
Standard Deviation ◽  
Real Time ◽  
Convolutional Neural Networks ◽  
Recurrent Neural Networks ◽  
Healthy Subjects ◽  
Hand Movements ◽  
Cord Injury ◽  
Field Programmable

AbstractThis paper reports on the evaluation of recurrent and convolutional neural networks as real-time grasp phase classifiers for future control of neuroprostheses for people with high spinal cord injury. A field-programmable gate array has been chosen as an implementation platform due to its form factor and ability to perform parallel computations, which are specific for the selected neural networks. Three different phases of two grasp patterns and the additional open hand pattern were predicted by means of surface Electromyography (EMG) signals (i.e. Seven classes in total). Across seven healthy subjects, CNN (Convolutional Neural Networks) and RNN (Recurrent Neural Networks) had a mean accuracy of 85.23% with a standard deviation of 4.77% and 112 µs per prediction and 83.30% with a standard deviation of 4.36% and 40 µs per prediction, respectively.

Determination of the Minimum Distance Between Moving Objects Including Velocity Information

1993 ◽  
Author(s):  
Meyer Nahon
Keyword(s):  
Minimum Distance ◽  
Moving Objects ◽  
Rapid Determination ◽  
Computation Time ◽  
Optimization Techniques ◽  
Real Time Control ◽  
Time Step ◽  
Time Control ◽  
Velocity Information

Abstract The rapid determination of the minimum distance between objects is of importance in collision avoidance for a robot maneuvering among obstacles. Currently, the fastest algorithms for the solution of this problem are based on the use of optimization techniques to minimize a distance function. Furthermore, to date this problem has been approached purely through the position kinematics of the two objects. However, although the minimum distance between two objects can be found quickly on state-of-the-art hardware, the modelling of realistic scenes entails the determination of the minimum distances between large numbers of pairs of objects, and the computation time to calculate the overall minimum distance between any two objects is significant, and introduces a delay which has serious repercussions on the real-time control of the robot. This paper presents a technique to modify the original optimization problem in order to include velocity information. In effect, the minimum distance calculation is performed at a future time step by projecting the effect of present velocity. This method has proven to give good results on a 6-dof robot maneuvering among obstacles, and has allowed a complete compensation of the lags incurred due to computational delays.

Accelerating Climate Model Computation by Neural Networks: A Comparative Study

2021 ◽  
Author(s):  
Maha Mdini ◽  
Takemasa Miyoshi ◽  
Shigenori Otsuka
Keyword(s):  
Neural Networks ◽  
Computational Complexity ◽  
Statistical Model ◽  
Physical Model ◽  
Climate Model ◽  
Numerical Models ◽  
Computation Time ◽  
High Accuracy ◽  
Physical Models ◽  
Computer Resources

<p>In the era of modern science, scientists have developed numerical models to predict and understand the weather and ocean phenomena based on fluid dynamics. While these models have shown high accuracy at kilometer scales, they are operated with massive computer resources because of their computational complexity.  In recent years, new approaches to solve these models based on machine learning have been put forward. The results suggested that it be possible to reduce the computational complexity by Neural Networks (NNs) instead of classical numerical simulations. In this project, we aim to shed light upon different ways to accelerating physical models using NNs. We test two approaches: Data-Driven Statistical Model (DDSM) and Hybrid Physical-Statistical Model (HPSM) and compare their performance to the classical Process-Driven Physical Model (PDPM). DDSM emulates the physical model by a NN. The HPSM, also known as super-resolution, uses a low-resolution version of the physical model and maps its outputs to the original high-resolution domain via a NN. To evaluate these two methods, we measured their accuracy and their computation time. Our results of idealized experiments with a quasi-geostrophic model [SO3] show that HPSM reduces the computation time by a factor of 3 and it is capable to predict the output of the physical model at high accuracy up to 9.25 days. The DDSM, however, reduces the computation time by a factor of 4 and can predict the physical model output with an acceptable accuracy only within 2 days. These first results are promising and imply the possibility of bringing complex physical models into real time systems with lower-cost computer resources in the future.</p>

Sign in / Sign up

Export Citation Format

Share Document