scholarly journals An SMT-Based Approach for Verifying Binarized Neural Networks

2021 ◽  
pp. 203-222
Author(s):  
Guy Amir ◽  
Haoze Wu ◽  
Clark Barrett ◽  
Guy Katz
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Formal Verification ◽  
Software Systems ◽  
Network Architectures ◽  
Security Issues ◽  
The Neural Network ◽  
Network Verification ◽  
Promising Avenue

AbstractDeep learning has emerged as an effective approach for creating modern software systems, with neural networks often surpassing hand-crafted systems. Unfortunately, neural networks are known to suffer from various safety and security issues. Formal verification is a promising avenue for tackling this difficulty, by formally certifying that networks are correct. We propose an SMT-based technique for verifying binarized neural networks — a popular kind of neural network, where some weights have been binarized in order to render the neural network more memory and energy efficient, and quicker to evaluate. One novelty of our technique is that it allows the verification of neural networks that include both binarized and non-binarized components. Neural network verification is computationally very difficult, and so we propose here various optimizations, integrated into our SMT procedure as deduction steps, as well as an approach for parallelizing verification queries. We implement our technique as an extension to the Marabou framework, and use it to evaluate the approach on popular binarized neural network architectures.

scholarly journals Increasing of Thermal Images Resolution Using Deep Learning Neural Networks

2021 ◽  
Vol 25 (3) ◽  
pp. 31-35
Author(s):  
Piotr Więcek ◽  
Dominik Sankowski
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Execution Time ◽  
High Accuracy ◽  
New Method ◽  
Residual Network ◽  
Thermal Images ◽  
The Neural Network

The article presents a new algorithm for increasing the resolution of thermal images. For this purpose, the residual network was integrated with the Kernel-Sharing Atrous Convolution (KSAC) image sub-sampling module. A significant reduction in the algorithm’s complexity and shortening the execution time while maintaining high accuracy were achieved. The neural network has been implemented in the PyTorch environment. The results of the proposed new method of increasing the resolution of thermal images with sizes 32 × 24, 160 × 120 and 640 × 480 for scales up to 6 are presented.

scholarly journals Convolutional neural network for image classification based on transfer learning technique

2021 ◽  
Author(s):  
Ghassan Mohammed Halawani
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Convolutional Neural Network ◽  
Image Classification ◽  
Transfer Learning ◽  
Learning Networks ◽  
The Neural Network ◽  
Learning Technique ◽  
Common Architecture

The main purpose of this project is to modify a convolutional neural network for image classification, based on a deep-learning framework. A transfer learning technique is used by the MATLAB interface to Alex-Net to train and modify the parameters in the last two fully connected layers of Alex-Net with a new dataset to perform classifications of thousands of images. First, the general common architecture of most neural networks and their benefits are presented. The mathematical models and the role of each part in the neural network are explained in detail. Second, different neural networks are studied in terms of architecture, application, and the working method to highlight the strengths and weaknesses of each of neural network. The final part conducts a detailed study on one of the most powerful deep-learning networks in image classification – i.e. the convolutional neural network – and how it can be modified to suit different classification tasks by using transfer learning technique in MATLAB.

scholarly journals GAIT RECOGNITION BASED ON CONVOLUTIONAL NEURAL NETWORKS

2017 ◽  
Vol XLII-2/W4 ◽  
pp. 207-212 ◽  
Author(s):  
A. Sokolova ◽  
A. Konushin
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Feature Extraction ◽  
Deep Learning ◽  
Deep Neural Network ◽  
Gait Recognition ◽  
Learning Approach ◽  
Network Architectures ◽  
Motion Information ◽  
Advantages And Disadvantages

In this work we investigate the problem of people recognition by their gait. For this task, we implement deep learning approach using the optical flow as the main source of motion information and combine neural feature extraction with the additional embedding of descriptors for representation improvement. In order to find the best heuristics, we compare several deep neural network architectures, learning and classification strategies. The experiments were made on two popular datasets for gait recognition, so we investigate their advantages and disadvantages and the transferability of considered methods.

scholarly journals An optical neural network using less than 1 photon per multiplication

2022 ◽  
Vol 13 (1) ◽  
Author(s):  
Tianyu Wang ◽  
Shi-Yuan Ma ◽  
Logan G. Wright ◽  
Tatsuhiro Onodera ◽  
Brian C. Richard ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Deep Neural Networks ◽  
Fundamental Principle ◽  
Energy Costs ◽  
Network Architectures ◽  
Optical Neural Networks ◽  
Optical Neural Network ◽  
Handwritten Digit

AbstractDeep learning has become a widespread tool in both science and industry. However, continued progress is hampered by the rapid growth in energy costs of ever-larger deep neural networks. Optical neural networks provide a potential means to solve the energy-cost problem faced by deep learning. Here, we experimentally demonstrate an optical neural network based on optical dot products that achieves 99% accuracy on handwritten-digit classification using ~3.1 detected photons per weight multiplication and ~90% accuracy using ~0.66 photons (~2.5 × 10−19 J of optical energy) per weight multiplication. The fundamental principle enabling our sub-photon-per-multiplication demonstration—noise reduction from the accumulation of scalar multiplications in dot-product sums—is applicable to many different optical-neural-network architectures. Our work shows that optical neural networks can achieve accurate results using extremely low optical energies.

scholarly journals Exploring deep neural networks via layer-peeled model: Minority collapse in imbalanced training

2021 ◽  
Vol 118 (43) ◽  
pp. e2103091118
Author(s):  
Cong Fang ◽  
Hangfeng He ◽  
Qi Long ◽  
Weijie J. Su
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Deep Neural Networks ◽  
Model Minority ◽  
Tight Frame ◽  
Learning Models ◽  
The Neural Network ◽  
Long Time ◽  
Topmost Layer

In this paper, we introduce the Layer-Peeled Model, a nonconvex, yet analytically tractable, optimization program, in a quest to better understand deep neural networks that are trained for a sufficiently long time. As the name suggests, this model is derived by isolating the topmost layer from the remainder of the neural network, followed by imposing certain constraints separately on the two parts of the network. We demonstrate that the Layer-Peeled Model, albeit simple, inherits many characteristics of well-trained neural networks, thereby offering an effective tool for explaining and predicting common empirical patterns of deep-learning training. First, when working on class-balanced datasets, we prove that any solution to this model forms a simplex equiangular tight frame, which, in part, explains the recently discovered phenomenon of neural collapse [V. Papyan, X. Y. Han, D. L. Donoho, Proc. Natl. Acad. Sci. U.S.A. 117, 24652–24663 (2020)]. More importantly, when moving to the imbalanced case, our analysis of the Layer-Peeled Model reveals a hitherto-unknown phenomenon that we term Minority Collapse, which fundamentally limits the performance of deep-learning models on the minority classes. In addition, we use the Layer-Peeled Model to gain insights into how to mitigate Minority Collapse. Interestingly, this phenomenon is first predicted by the Layer-Peeled Model before being confirmed by our computational experiments.

scholarly journals Cinemática Inversa com Redes Neurais Aplicadas em Robôs Manipuladores

2018 ◽  
Vol 1 (43) ◽  
pp. 49
Author(s):  
Gabriel Daltro Duarte ◽  
Claudio Pereira Mego Quinteros ◽  
Lincoln Machado Araújo
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Nonlinear Equations ◽  
Inverse Kinematics ◽  
Robot Manipulator ◽  
Kinematic Model ◽  
Online Training ◽  
Network Architectures ◽  
The Neural Network ◽  
Neural Network Architectures

Today’s world is going through what is known as the Fourth Industrial Revolution. Robots have been gaining more and more space in the industry and going beyond expectations. The use of robots in industry is related to the increasing production and the quality of the electronic products. For an accurate movement of a robot manipulator it is necessary to obtain its inverse kinematic model, however, obtaining this model requires the challenging solution of a set of nonlinear equations. For this system of nonlinear equations there is no generic solution method. In view of this matter, this research aims to solve the problem of the inverse kinematics of a robot manipulator using artificial neural networks without the need to model the robot’s direct kinematics. Two neural network training methodologies, called offline training and online training were used. The basic difference between these two is that in the offline methodology all training points are obtained before any training of the neural network occurs, whereas in online training the use of a training method is recurrent. As the robot moves, new training points are obtained and training processes with the new acquired points are executed, allowing a learning process of continuous inverse kinematics. To validate the proposed methodology a prototype of a manipulated planar robot with one degree of freedom was developed and several architectures of neural networks were tested to find the optimal architecture. The offline training methodology obtained very satisfactory results for most of the neural network architectures tested. The online training only achieved satisfactory results in neural network architectures with quantities of neurons much larger than the quantities used in the architectures used in offline training and still obtained inferior results. The neural networks trained in offline mode, when compared to the training networks in the online mode, presented a greater capacity of generalization and a smaller value of output error. The online training has only achieved satisfactory results in neural network architectures with quantities of neurons much larger than the quantities used in the trained architectures in offline mode.

scholarly journals Solving Differential Equations Using Feedforward Neural Networks

2021 ◽  
pp. 385-399
Author(s):  
Wilson Guasti Junior ◽  
Isaac P. Santos
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Differential Equations ◽  
Feedforward Neural Networks ◽  
Activation Functions ◽  
Learning Models ◽  
Python Language ◽  
The Neural Network ◽  
Minimization Methods

Abstract In this work we explore the use of deep learning models based on deep feedforward neural networks to solve ordinary and partial differential equations. The illustration of this methodology is given by solving a variety of initial and boundary value problems. The numerical results, obtained based on different feedforward neural networks structures, activation functions and minimization methods, were compared to each other and to the exact solutions. The neural network was implemented using the Python language, with the Tensorflow library.

scholarly journals Automatic Defects Segmentation and Identification by Deep Learning Algorithm with Pulsed Thermography: Synthetic and Experimental Data

2021 ◽  
Vol 5 (1) ◽  
pp. 9
Author(s):  
Qiang Fang ◽  
Clemente Ibarra-Castanedo ◽  
Xavier Maldague
Keyword(s):  
Neural Network ◽  
Experimental Data ◽  
Neural Networks ◽  
Deep Learning ◽  
Infrared Thermography ◽  
Learning Algorithm ◽  
Synthetic Data ◽  
Training Data ◽  
Experimental Database ◽  
The Neural Network

In quality evaluation (QE) of the industrial production field, infrared thermography (IRT) is one of the most crucial techniques used for evaluating composite materials due to the properties of low cost, fast inspection of large surfaces, and safety. The application of deep neural networks tends to be a prominent direction in IRT Non-Destructive Testing (NDT). During the training of the neural network, the Achilles heel is the necessity of a large database. The collection of huge amounts of training data is the high expense task. In NDT with deep learning, synthetic data contributing to training in infrared thermography remains relatively unexplored. In this paper, synthetic data from the standard Finite Element Models are combined with experimental data to build repositories with Mask Region based Convolutional Neural Networks (Mask-RCNN) to strengthen the neural network, learning the essential features of objects of interest and achieving defect segmentation automatically. These results indicate the possibility of adapting inexpensive synthetic data merging with a certain amount of the experimental database for training the neural networks in order to achieve the compelling performance from a limited collection of the annotated experimental data of a real-world practical thermography experiment.

scholarly journals Book spine recognition with the use of deep neural networks

2020 ◽  
Vol 44 (6) ◽  
pp. 968-977
Author(s):  
M.O. Kalinina ◽  
P.L. Nikolaev
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Network Performance ◽  
Deep Neural Network ◽  
Deep Neural Networks ◽  
Original Data ◽  
Network Architectures ◽  
Data Set ◽  
The Neural Network ◽  
Processing Information

Nowadays deep neural networks play a significant part in various fields of human activity. Especially they benefit spheres dealing with large amounts of data and lengthy operations on obtaining and processing information from the visual environment. This article deals with the development of a convolutional neural network based on the YOLO architecture, intended for real-time book recognition. The creation of an original data set and the training of the deep neural network are described. The structure of the neural network obtained is presented and the most frequently used metrics for estimating the quality of the network performance are considered. A brief review of the existing types of neural network architectures is also made. YOLO architecture possesses a number of advantages that allow it to successfully compete with other models and make it the most suitable variant for creating an object detection network since it enables some of the common disadvantages of such networks to be significantly mitigated (such as recognition of similarly looking, same-color book coves or slanted books). The results obtained in the course of training the deep neural network allow us to use it as a basis for the development of the software for book spine recognition.

Sign in / Sign up

Export Citation Format

Share Document