Loading...
Search for:
deep-neural-networks
0.007 seconds
Total 146 records
Inverse design of compact power divider with arbitrary outputs for 5G applications
, Article Scientific Reports ; Volume 12, Issue 1 , 2022 ; 20452322 (ISSN) ; Tavakol, M. R ; Atlasbaf, Z ; Sharif University of Technology
Nature Research
2022
Abstract
Since the recent on-demand applications need more sophisticated circuits and subsystems, components with configurable capabilities attract attention more than before in commercial systems, specifically the fifth generation (5G). Power dividers play a crucial role in 5G phased array systems, and their role becomes more significant if the output powers ratio is adjustable. Here, we suggest a design methodology by which planar power splitters with arbitrary output power levels can be designed in light of very simple perturbations, i.e., vias. Through our design procedure, we find an optimized pattern for hybrid vias-some of them are made of PEC, and others are dielectric, e.g., air,...
Utilization of Different Optical Wavelengths in Diffractive Deep Neural Networks for Object Classification in Multi-Channel Images
, M.Sc. Thesis Sharif University of Technology ; Vosughi Vahdat, Bijan (Supervisor) ; Kavehvash, Zahra (Supervisor)
Abstract
Diffractive deep neural network is an optical machine-learning framework that uses diffractive surfaces, optical devices, electro-optic devices and engineered matterials to optically perform computational tasks. These diffractive networks, after their desing and train phase by computers and machine learning algorithms, are physically fabricated using 3D printing or lithography, to actualize the model of trained network. Machine learning processes and alghorithms are performed through light-matter interaction and diffraction of light. This procedure is done at the speed of light and without the need of any power, except for the light illumination for the input object. In comparison with...
A Fast and Scalable Network-on-Chip for DNN Accelerators
, M.Sc. Thesis Sharif University of Technology ; Sarbazi Azad, Hamid (Supervisor)
Abstract
Deep Neural Networks (DNNs) are widely used as a promising machine learning method in different applications and come with intensive computation and storage requirements. In recent years, several pieces of prior work have proposed different accelerators to improve DNNs processing. We observe that although the state-of-the-art DNN accelerators effectively process some network layers of certain shapes, they fail to keep computation resources fully utilized for many other layers. The reason is twofold: first, the mapping algorithm is unable to employ all compute cores for processing some layer types and dimension sizes, and second, the hardware cannot perform data distribution and aggregation...
Using Deep Neural Networks in Reinforcement Learning
, M.Sc. Thesis Sharif University of Technology ; Soleymani Baghshah, Mahdieh (Supervisor) ; Rabiei, Hamidreza (Supervisor)
Abstract
Reinforcement learning is a field of machine learning which is more similar to human training procedures.It uses reward signals to train an agent designed to act in that environment. Deep neural networks enhance the agent’s ability to determine and act better in its complex environment. Most previous works have addressed model-free agents, which ignore modeling details of the environment that in turn can be used to achieve better results. On the other hand, humans utilize a model-based approach in their decision-making process. They use their knowledge to predict the future and choose the action that leads them to a better state. To combine the benefits of model-based and model-free designs,...
Adversarial Attacks on Deep Neural Networks
, Ph.D. Dissertation Sharif University of Technology ; Kasaei, Shohreh (Supervisor)
Abstract
The remarkable progress of deep neural networks in recent years has led to their entry into the industry and their use in the real world. However, one of the most important and basic issues that threaten the security of these networks is attacks. The attacks that deliberately manipulate input data cause vulnerabilities and misclassify networks. Due to the wide range of ways in which attacks can perturb input data, identifying their types is considered a vital part of ensuring a robust network. The inability of deep networks to generalize to unseen data is also an important limitation. This thesis presents a 2D adversarial attack and a 3D defense in this regard.In 2D attacks, the type of...
Design and Evaluation of a Reconfigurable Accelerator for Sparse Neural Networks
, M.Sc. Thesis Sharif University of Technology ; Sarbazi Azad, Hamid (Supervisor)
Abstract
Deep Neural Networks (DNNs) are widly used in various domains, such as medicine, engineering, industry, financial markets, mathematics and management. DNNs are composed of several layers, such as convolutional and fully connected layers. Increasing the number of layers in DNNs provides different application with their required accuracy. In recent years, there have been many accelerator aiming to execute DNNs. However, the high computation and memory demands in DNNs are the main challenges to execute DNNs. To reduce computation and memory requirements, various methods such as pruning and quantization, have been proposed. Pruning and quantization methods make the DNNs sparse and increase the...
A machine learning framework for predicting entrapment efficiency in niosomal particles
, Article International Journal of Pharmaceutics ; Volume 627 , 2022 ; 03785173 (ISSN) ; Aftab, A ; Ghaemmaghami, S ; Sharif University of Technology
Elsevier B.V
2022
Abstract
Niosomes are vesicles formed mostly by nonionic surfactant and cholesterol incorporation as an excipient. The drug entrapment efficiency of niosomal vesicles is particularly important and depends on many parameters. Changing the effective parameters to have maximum entrapment efficiency in the laboratory is time-consuming and costly. In this study, a machine learning framework was proposed to address these problems. In order to find the most critical parameter affecting the entrapment efficiency and its optimal value in a specific experiment, data were first extracted from articles of the last decade using keywords of niosome and thin-film hydration method. Then, deep neural network (DNN),...
A deep learning approach for the solution of probability density evolution of stochastic systems
, Article Structural Safety ; Volume 99 , 2022 ; 01674730 (ISSN) ; Khodabakhsh, A. H ; Sharif University of Technology
Elsevier B.V
2022
Abstract
Derivation of the probability density evolution provides invaluable insight into the behavior of many stochastic systems and their performance. However, for most real-time applications, numerical determination of the probability density evolution is a formidable task. The latter is due to the required temporal and spatial discretization schemes that render most computational solutions prohibitive and impractical. In this respect, the development of an efficient computational surrogate model is of paramount importance. Recent studies on the physics-constrained networks show that a suitable surrogate can be achieved by encoding the physical insight into a deep neural network. To this aim, the...
Code Summarization in Event-Driven Programs
, M.Sc. Thesis Sharif University of Technology ; Heydarnoori, Abbas (Supervisor)
Abstract
Developers have been spending a lot of time on program comprehension during software evolution. Program comprehension reduces the cost and the software development time and increases maintainability of a program. However, the lack of documentation makes this process exhausting. Source code summarization is one of the existing solutions to help developers understand a program. Source code summarization gives an opportunity for the developers to better understand the source code by spending less time. There are a lot of approaches for source code summarization. For instance, exploiting knowledge of the crowd, information retrieval, deep neural networks, or using eye tracking of developers...
Mathematical Foundations of Deep Learning: a Theoretical Framework for Generalization
, M.Sc. Thesis Sharif University of Technology ; Alishahi, Kasra (Supervisor) ; Hadji Mirsadeghi, Mir Omid (Co-Supervisor)
Abstract
Deep Neural Networks, are predictive models in Machine Learning, that during the last decade they've had a great success. However being in an over-parametrized and highly non-convex regime, the analytical examinations of these models is quite a challenging task to do. The empirical developments of Neural Networks, and their distinguishing performance in prediction problems, has motivated researchers, to formalize a theoretical foundations for these models and provide us with a framework, in which one can explain and justify their behavior and properties. this framework is of great importance because it would help us to come to a better understanding of how these models work and also enables...
User-Centric Recommendation for Mobile Notification Servicees
, M.Sc. Thesis Sharif University of Technology ; Soleymani Baghshah, Mahdieh (Supervisor)
Abstract
With the popularity of smart devices, a lot of applications have developed and deployed.Developers try to establish continuous interaction with their users by different tools including push notifications. Push notification is a message that is sent from developers to the users and as soon as the user’s device receives that, it appears on the device screen. Sending proper content to users in order to resume their engagement is one of the most important usages of notifications. Users are not interested in receiving irrelevant notifications, and receiving irrelevant notifications make them remove the application, so it’s important to predict users’ interest in different notifications and push...
Deep Learning Algorithms for Solving Graph Problems
, M.Sc. Thesis Sharif University of Technology ; Salehkaleybar, Saber (Supervisor) ; Hashemi, Matin (Co-Supervisor)
Abstract
Nowadays, thanks to improvement of processing hardware and plenty of data available, artificial intelligence and specifically Deep Learning are being one of the powerful tools for solving different problems. Also graph is one of the powerful tools for modeling different data structures. Graph matching is one of the problems in the field of graph problems.In this thesis we consider the problem of graph matching in Erdos-Renyi graphs. The graph matching problem refers to recovering the node-to-node correspondence between two correlated graphs. Previous works theoretically showed that recovering is feasible in sparse Erdos-R´enyi graphs if and only if the probability of having an edge between a...
Fake News Spreading Mitigation Via Tracing Information In Social Networks
, M.Sc. Thesis Sharif University of Technology ; Rabiee, Hamid Reza (Supervisor)
Abstract
Expansion of using social networks for reading and sharing news on the one hand, allows for quick and easy access to the news, but on the other hand, increases the spread of fake news and misleads many users. Therefore, identifying the fake news and attempting to mitigate its spreading on social networks is one of the major research areas in recent years. An important issue in this regard is reducing the time gap between news release time and identifying it as fake and starting to take action. Many researches has been done on detection of fake news, but since there is a trade-off between minimizing the time gap and maximizing accuracy, work on these models continues. There are also different...
Designing an Intelligent System to Analyze Electrograms of Induced Pluripotent Stem Cell-Derived Cardiomyocytes
, M.Sc. Thesis Sharif University of Technology ; Rabiee, Hamid Reza (Supervisor) ; Soleymani, Mahdieh (Supervisor) ; Pahlavan, Sara (Co-Advisor)
Abstract
Ability to differentiate induced pluripotent stem cells to cardiomycocytes has attracted attentions,considering crucial role of the heart in the human body and great potential applications of these cells like disease modeling, new treatment methods and basic research. We are able to analyze the performance of beating cells through recording extracellular field potentials of cardiomyocytes using multi-electrode array (MEA) technology. This analysis is an essential step to use cardiac cells in any future development and experiment. Currently, the electrophysiology experts analyze recorded extracellular field potentials of induced cardiomyocytes by observing all the episodes of each record....
Adversarial Robustness of Deep Neural Networks in Text Domain
, M.Sc. Thesis Sharif University of Technology ; Soleymani Baghshah, Mahdieh (Supervisor)
Abstract
In recent years, neural networks have been widely used in most machine learning domains. However, it has been shown that these networks are vulnerable to adversarial examples. adversarial examples are small and imperceptible perturbations applied to the input which lead to producing wrong output and thus, fooling the network. This will become an important issue in security related applications of deep neural networks, such as self-driving cars and medical diagnostics. Since, in the wort-case scenario, even human lives could be threatened. Although, many works have focused on crafting adversarial examples for image data, only a few studies have been done on textual data due to the existing...
Exploiting Transfer Learning in Deep Neural Networks for Time Series
, M.Sc. Thesis Sharif University of Technology ; Manzuri Shalmani, Mohammad Taghi (Supervisor)
Abstract
The importance of transfer learning in image-related problems comes from its many advantages that are sometimes undeniable. Previous researches have well shown the success of transfer learning in this area using deep neural networks. However, transfer learning for time series data has not yet been done in a conventional and automated manner. The main reason for avoiding transfer learning in this domain relates to the dynamic and stochastic nature of the time series, where they show a time-varying behavior. Previous experiments have shown that transfer learning between two heterogeneous time series could harm the forecasting accuracy of a model. Therefore, in this thesis, we aim to explore...
Weakly Supervised Semantic Segmentation Using Deep Neural Networks
, M.Sc. Thesis Sharif University of Technology ; Kasaei, Shohreh (Supervisor)
Abstract
Semantic segmentation which is the classification of every pixel in an input image is a fundamental task in the fields of computer vision and scene understanding. Applications of semantic segmentation include usage in autonomous vehicles and robotics. Since in this task dense annotation of images in the dataset is needed, recent methods have been proposed to utilize weakly-supervised and semi-supervised learning using data with weak labels and unlabeled data respectively. Because the amount of fully labeled data might not be sufficient in such methods, some papers have proposed to employ depth input data due to its rich geometrical and local information when available. In this research, an...
Adversarial Robustness Against Perceptual and Unforeseen Attacks in Deep Neural Networks in Images
, M.Sc. Thesis Sharif University of Technology ; Rohban, Mohammad Hossein (Supervisor)
Abstract
Improvements in deep neural networks and their widespread use in research and practical application have raised significant concerns about the robustness of these networks against adversarial examples designed to deceive a deep network in calcu- lating the correct output with a slight change in input. Since this is an essential issue in highly sensitive applications, it is necessary to use a training method that reduces the model’s sensitivity to these changes but still preserves the accuracy. The common method for this goal is training the model with adversarial examples. In other words, adversarial examples are generated during the training, and the model is trained with them. This allows...
Efficient Hardware-based Implementation of Object Detection in mmW-Imaging Systems Using AI Algorithms
, M.Sc. Thesis Sharif University of Technology ; Shabany, Mahdi (Supervisor) ; Kavehvash, Zahra (Supervisor)
Abstract
Today, due to the increasing activity of terrorist groups, monitoring people in important and busy places such as airports and train stations is very important. One of the technologies that has been developed for this purpose in recent years is 3D imaging technology using millimeter wave. These systems use millimeter waves to image people and identify objects hidden under clothing, which do not have the limitations of conventional imaging techniques such as x-rays and metal detectors. One of the advantages of using these systems is the ability to automatically detect objects in millimeter wave images using deep neural networks such as Segmented, Faster R-CNN and Mask R-CNN, which using these...
Investigation, Design and Simulation of Elastography and IOTA Imaging Modes in Ultrasound Systems
, M.Sc. Thesis Sharif University of Technology ; Vosoughi Vahdat, Bijan (Supervisor) ; Kavehvash, Zahra (Supervisor)
Abstract
Ultrasound is one of the most widely used imaging techniques used by physicians and radiologists as a tool. It has many diagnostic and therapeutic uses. In particular, ultrasound imaging is used for prenatal imaging due to relative safety, low cost, nature Non-ionizing, instant display and ease of use by the operator, widely used in many countries around the world. According to the Ministry of Health, ovarian cancer is the most important cause of death in women The effect of cancer in Iran. Breast cancer is also one of the most common cancers in women, followed by cancer Cervix and ovarian and uterine cancers are also among the most common cancers in women. This issue is not only in Iran It...