Digital Commons Network^™

Exploring Virtual Worlds With Cultural Algorithms: Ancient Alpena-Amberley Land Bridge, Thomas Joseph Palazzolo

Wayne State University Dissertations

In this thesis the Land Bridge system (DEEPDIVE) is described. The goal of the project is to use Artificial Intelligence technology to aid Archaeologists in the discovery of ancient prehistoric sites, now underwater. The example used here is the Alpena-Amberley Land Bridge that stretched across Lake Huron from Alpena in Michigan to Amberley in Ontario. During the Ice Age (around 10,000 years ago) it was above water for several thousand years. It was postulated that during that time it was used as a migration pathway for caribou, a major food source then. AI techniques were used to create a virtual …

Coverage Guided Differential Adversarial Testing Of Deep Learning Systems, Jianmin Guo, Houbing Song, Yue Zhao, Yu Jiang

Publications

Deep learning is increasingly applied to safety-critical application domains such as autonomous cars and medical devices. It is of significant importance to ensure their reliability and robustness. In this paper, we propose DLFuzz, the coverage guided differential adversarial testing framework to guide deep learing systems exposing incorrect behaviors. DLFuzz keeps minutely mutating the input to maximize the neuron coverage and the prediction difference between the original input and the mutated input, without manual labeling effort or cross-referencing oracles from other systems with the same functionality. We also design multiple novel strategies for neuron selection to improve the neuron coverage. The …

Towards Robust Artificial Intelligence Systems, Sunny Raj

Electronic Theses and Dissertations, 2020-2023

Adoption of deep neural networks (DNNs) into safety-critical and high-assurance systems has been hindered by the inability of DNNs to handle adversarial and out-of-distribution input. State-of-the-art DNNs misclassify adversarial input and give high confidence output for out-of-distribution input. We attempt to solve this problem by employing two approaches, first, by detecting adversarial input and, second, by developing a confidence metric that can indicate when a DNN system has reached its limits and is not performing to the desired specifications. The effectiveness of our method at detecting adversarial input is demonstrated against the popular DeepFool adversarial image generation method. On a …

Fast Decision-Making Under Time And Resource Constraints, Kyle Gabriel Lassak

Graduate Theses, Dissertations, and Problem Reports (ETD)

Practical decision makers are inherently limited by computational and memory resources as well as the time available in which to make decisions. To cope with these limitations, humans actively seek methods which limit their resource demands by exploiting structure within the environment and exploiting a coupling between their sensing and actuation to form heuristics for fast decision-making. To date, such behavior has not been replicated in artificial agents. This research explores how heuristics may be incorporated into the decision-making process to quickly make high-quality decisions through the analysis of a prominent case study: the outfielder problem. In the outfielder problem, …

Representation Learning With Adversarial Latent Autoencoders, Stanislav Pidhorskyi M.S.

Graduate Theses, Dissertations, and Problem Reports (ETD)

A large number of deep learning methods applied to computer vision problems require encoder-decoder maps. These methods include, but are not limited to, self-representation learning, generalization, few-shot learning, and novelty detection. Encoder-decoder maps are also useful for photo manipulation, photo editing, superresolution, etc. Encoder-decoder maps are typically learned using autoencoder networks.
Traditionally, autoencoder reciprocity is achieved in the image-space using pixel-wise
similarity loss, which has a widely known flaw of producing non-realistic reconstructions. This flaw is typical for the Variational Autoencoder (VAE) family and is not only limited to pixel-wise similarity losses, but is common to all methods relying upon …

Image And Video-Based Autism Spectrum Disorder Detection Via Deep Learning, Mindi Ruan

Graduate Theses, Dissertations, and Problem Reports (ETD)

People with Autism Spectrum Disorder (ASD) show atypical attention to social stimuli and aberrant gaze when viewing images of the physical world. However, it is unknown how they perceive the world from a first-person perspective. In this study, we used machine learning to classify photos taken in three different categories (people, indoors, and outdoors) as either having been taken by individuals with ASD or by peers without ASD. Our classifier effectively discriminated photos from all three categories but was particularly successful at classifying photos of people with >80% accuracy. Importantly, the visualization of our model revealed critical features that led …

Searches For Fast Radio Bursts Using Machine Learning, Devansh Agarwal

Graduate Theses, Dissertations, and Problem Reports (ETD)

Fast Radio bursts (FRBs) are enigmatic astrophysical events with millisecond durations and flux densities in the range 0.1-100 Jy, with the prototype source discovered by Lorimer et al. (2007). Like pulsars, FRBs show the characteristic inverse square sweep in observing frequency due to propagation through an ionized medium. This effect is quantified by the dispersion measure (DM). Unlike pulsars, FRBs have anomalously high DMs, which are consistent with an extragalactic origin. Over 100 FRBs have been published at the time of writing, and 13 have been conclusively identified with host galaxies with spectroscopically determined redshifts in the range 0.003 ≤ …

Human Path Prediction Using Auto Encoder Lstms And Single Temporal Encoders, Hayden Hudgins

Master's Theses

Due to automation, the world is changing at a rapid pace. Autonomous agents have become more common over the last several years and, as a result, have created a need for improved software to back them up. The most important aspect of this greater software is path prediction, as robots need to be able to decide where to move in the future. In order to accomplish this, a robot must know how to avoid humans, putting frame prediction at the core of many modern day solutions. A popular way to solve this complex problem of frame prediction is Auto Encoder …

Sparsity And Weak Supervision In Quantum Machine Learning, Seyran Saeedi

Theses and Dissertations

Quantum computing is an interdisciplinary field at the intersection of computer science, mathematics, and physics that studies information processing tasks on a quantum computer. A quantum computer is a device whose operations are governed by the laws of quantum mechanics. As building quantum computers is nearing the era of commercialization and quantum supremacy, it is essential to think of potential applications that we might benefit from. Among many applications of quantum computation, one of the emerging fields is quantum machine learning. We focus on predictive models for binary classification and variants of Support Vector Machines that we expect to be …

Red Blood Cell Segmentation And Classification From Microscopic Images Using Machine Learning, Korranat Naruenatthanaset

Chulalongkorn University Theses and Dissertations (Chula ETD)

Red blood cell morphology analysis plays an essential role in diagnosing many diseases caused by RBC disorders. This manual inspection is a long process and requires practice and experience. Since recent computer vision and image processing in the medical imaging area can provide efficient tools, it can help hematologists to automatically analyze images from a microscope in a reduced time and cost. This research presents a new method to segment and classify RBCs from blood smear images. The process started from data collection, which a new application was created for precisely labeling. The normalization was done to reduce the color …

A Robust System For Core Thai Natural Language Processing Technologies, Can Udomcharoenchaikit

Chulalongkorn University Theses and Dissertations (Chula ETD)

As the amount of unstructured textual data grows, it becomes increasingly important to build an intelligent system that can process it. Natural Language Processing (NLP) is a technology that allows a computer to exploit human languages to perform tasks. Deep learning models have shown excellent results across fundamental tasks in NLP, such as word segmentation, part-of-speech tagging, and named-entity recognition. However, in many situations, these proposed methods fail to perform well. For an NLP system to be robust, it must address issues such as out-of-vocabulary and spelling-mistakes. This thesis's research goal is to develop NLP models that can handle malformed …

Semi-Supervised Thai Sentence Segmentation Using Local And Distant Word Representations, Chanatip Saetia

Chulalongkorn University Theses and Dissertations (Chula ETD)

A sentence is typically treated as the minimal syntactic unit used for extracting valuable information from a longer piece of text. However, in written Thai, there are no explicit sentence markers. We proposed a deep learning model for the task of sentence segmentation that includes three main contributions. First, we integrate n-gram embedding as a local representation to capture word groups near sentence boundaries. Second, to focus on the keywords of dependent clauses, we combine the model with a distant representation obtained from self-attention modules. Finally, due to the scarcity of labeled data, for which annotation is difficult and time-consuming, …

Deep Sequential Real Estate Recommendation Approach For Solving Item Cold Start Problem, Jirut Polohakul

Chulalongkorn University Theses and Dissertations (Chula ETD)

The item cold-start problem occurs when a recommendation system cannot recommend new items owing to record deficiencies and new listing omissions. When searching for real estate, users can register a concurrent interest in recent and prior projects. Thus, an approach to recommend cold-start and warm-start items simultaneously must be determined. Furthermore, unrequired membership and stop-by behavior cause real estate recommendations to have many cold-start and new users. This characteristic encourages the use of a content-based approach and a session-based recommendation system. Herein, we propose a real estate recommendation approach for solving the item cold-start problem with acceptable warm-start item recommendations …

A Real Estate Valuation Model Using Boosted Feature Selection, Kankawee Chanasit

Chulalongkorn University Theses and Dissertations (Chula ETD)

To estimate real estate values, a complex valuation model based on artificial neural network (ANN) has been established as a successful means in modern machine learning research, specifically when high-dimensional data are available. Unfortunately, the real estate data in many locations, such as Thailand, are quite limited in terms of features. Hence, it becomes mandatory to reduce the complexity using feature selection techniques. These techniques aim to improve performance by identifying significant factors and help decrease the computational overload and model construction. However, due to the lack of explicability and interpretability in ANNs, the analysis of input factors cannot be …

Detection Of Wagyu Beef Sources With Image Classification Using Convolutional Neural Network, Nattakorn Kointarangkul

Chulalongkorn University Theses and Dissertations (Chula ETD)

Wagyu beef originated in Japan. However, there are many types of Wagyu beef in the market around the globe. Primary sources include Australia, USA, Canada and the United Kingdom. The authentic Japanese Wagyu is well known for its intense marbling, juicy rich flavor and tenderness. Observing that there are differences in flavor, texture, and quality between distinct sources of Wagyu. This research presents an AI-based approach to identify Wagyu beef sources with image classification. The input images were collected from reliable sources on the internet and augmented with DCGAN. Deep neural networks, CNN, was constructed to detect the marbled fat …

Using Automatic Speech Recognition To Assess Thai Speech Language Fluency In Montreal Cognitive Assessment (Moca), Pimarn Kantithammakorn

Chulalongkorn University Theses and Dissertations (Chula ETD)

The Montreal Cognitive Assessment (MoCA), a widely accepted screening tool for identifying patients with mild cognitive impairment (MCI), includes a language fluency test of verbal functioning where scores are based on the number of unique correct words produced by the test-taker. However, with different languages, it is possible that unique words may be counted differently. This study focuses on Thai as a language that differs from English in its type of word combination. We applied various automatic speech recognition (ASR) techniques to develop an assisted scoring system for the language fluency test of the MoCA with Thai language support. The …

Accurate Surface Ultraviolet Radiation Forecasting For Clinical Applications With Deep Neural Network, Raksit Raksasat

Chulalongkorn University Theses and Dissertations (Chula ETD)

Exposure to appropriate doses of UV radiation provides enormously health and medical treatment benefits including psoriasis. Typical hospital-based phototherapy cabinets contain a bunch of artificial lamps, either broad-band (main emission spectrum 280-360 nm, maximum 320 nm), or narrow-band UV B irradiation (main emission spectrum 310-315nm, maximum 311nm). For patients who cannot access phototherapy centers, sun-bathing, or heliotherapy, can be a safe and effective treatment alternative. However, as sunlight contains the full range of UV radiation (290-400 nm), careful sun-bathing supervised by photodermatologist based on accurate UV radiation forecast is vital to minimize potential adverse effects. Here, using 10-year UV radiation …

Multi-Evidence Learning For Medical Diagnosis, Tongjai Yampaka

Chulalongkorn University Theses and Dissertations (Chula ETD)

In recent years, a great many approaches for learning from multiple sources by considering the diversity of different views have been proposed. The most interesting field is medical diagnosis. For example, breast cancer screening normally employs two views of mammography (Cranio-Caudal and Medio-Lateral-Oblique) or two modes of ultrasound (B-mode and Doppler mode) breast images. This study proposes a multi-evidence learning model that combines the multiple evidences of breast images to improve diagnosis. Two views mammography and two modes of ultrasound were used. Our proposed model consists of four stages. First, feature extraction using Convolutional Neuron Networks was operated to extract …

Artificial Intelligence And Copyright Law In Singapore A Study On The Protection Of Compilations And Databases Arranged By Ai-Systems, Sella Say

Chulalongkorn University Theses and Dissertations (Chula ETD)

While the capability of artificial intelligence ("AI") gains remarkable momentum in creating copyrightable materials – the questions regarding the eligibility of these new creations, at the moment, are broadly discussed and posed challenges to the regime. The problem of how we fit the conventional notion of authorship and the condition of originality for AI-generated works remains a controversial topic. Some might suggest that subject matter created by AI should not be granted copyright protection on the presumption that AI is not a human who could treat as authors of works. At the same time, other supportive claims that the first …

Attack Detection And Mitigation In Mobile Robot Formations, Arnold Fernandes

Masters Theses

"A formation of cheap and agile robots can be deployed for space, mining, patrolling, search and rescue applications due to reduced system and mission cost, redundancy, improved system accuracy, reconfigurability, and structural flexibility. However, the performance of the formation can be altered by an adversary. Therefore, this thesis investigates the effect of adversarial inputs or attacks on a nonholonomic leader-follower-based robot formation and introduces novel detection and mitigation schemes.

First, an observer is designed for each robot in the formation in order to estimate its state vector and to compute the control law. Based on the healthy operation of the …

Computer Vision Based Deep Learning Models For Cyber Physical Systems, Muhammad Monjurul Karim

Masters Theses

“Cyber-Physical Systems (CPSs) are complex systems that integrate physical systems with their counterpart cyber components to form a close loop solution. Due to the ability of deep learning in providing sensor data-based models for analyzing physical systems, it has received increased interest in the CPS community in recent years. However, developing vision data-based deep learning models for CPSs remains critical since the models heavily rely on intensive, tedious efforts of humans to annotate training data. Besides, most of the models have a high tradeoff between quality and computational cost. This research studies deep learning algorithms to achieve affordable and upgradable …

Real-Time Traffic Sign Detection And Classification Based On A Video Feed, Victor Ciuntu

Graduate Research Theses & Dissertations

Autonomous vehicle development is currently progressing at a very fast pace and traffic sign detection and classification has an important role in it. This thesis looks at the history of autonomous vehicles as well as different implementations for traffic sign detection and classification. Multiple possible approaches are analyzed with the final goal of doing this task in real-time using a portable system.

To accomplish this task, the final solution uses a convolutional neural network for detection and classification combined with a custom optical character recognition algorithm for speed limit signs. The optical character recognition algorithm is built from the ground …

An Automated Method For Detecting Water Levels Using Computer Vision And Artificial Intelligence, Priyanjani Chowdary Chandra

Graduate Research Theses & Dissertations

Flooding is one of the most dangerous weather events today. Between 2015-2019, on average, it has caused more than 130 deaths every year in the USA alone. World Health Organization has reported that, between 1998-2017, floods have affected more than 2 billion people worldwide. The devastating nature of flood necessitates the continuous monitoring of water level in the rivers and streams in flood-prone areas to detect the incoming flood. In this thesis, we have designed and implemented a computer vision and AI-based system that continuously detect the water level in the creek. Our solution employs an effective template matching algorithm …

Satellite Constellation Deployment And Management, Joseph Ryan Kopacz

Electronic Theses and Dissertations

This paper will review results and discuss a new method to address the deployment and management of a satellite constellation. The first two chapters will explorer the use of small satellites, and some of the advances in technology that have enabled small spacecraft to maintain modern performance requirements in incredibly small packages.

The third chapter will address the multiple-objective optimization problem for a global persistent coverage constellation of communications spacecraft in Low Earth Orbit. A genetic algorithm was implemented in MATLAB to explore the design space – 288 trillion possibilities – utilizing the Satellite Tool Kit (STK) software developers kit. …

Fusion-Net: Integration Of Dimension Reduction And Deep Learning Neural Network For Image Classification, Mohammad Masum, Philippe Laval

Published and Grey Literature from PhD Candidates

Building a deep network using original digital images requires learning many parameters which may reduce the accuracy rates. The images can be compressed by using dimension reduction methods and extracted reduced features can be feeding into a deep network for classification. Hence, in the training phase of the network, the number of parameters will be decreased. Principal Component Analysis is a well-known dimension reduction technique that leverage orthogonal linear transformation of the original data. In this paper, we propose a neural network-based framework, named Fusion-Net, which implements PCA on an image dataset (CIFAR-10) and then a neural network applies on …

Perceptions, Potholes, And Possibilities Of Using Digital Voice Assistants To Differentiate Instructions, Adrian A. Weir

Walden Dissertations and Doctoral Studies

Access to technologies and understanding the potential uses of technology to differentiate instruction have been a concern for the teachers and students in a local school district located in the southeastern United States. Despite the emergence of digital voice assistants (DVAs) as tools for instructions, teachers lack knowledge and strategies for using DVAs to differentiate instruction in their classrooms. The purpose of this qualitative study was to identify teacher knowledge and strategies employed among special education (SPED) teachers using DVAs to differentiate instruction in their classrooms. The concepts of Carol Tomlinson’s differentiation theory and Mishra and Koehler’s TPACK framework served …

Compound Vision Approach For Autonomous Vehicles Navigation, Michael Mikhael

Open Access Theses & Dissertations

An analogy can be made between the sensing that occurs in simple robots and drones and that in insects and crustaceans, especially in basic navigation requirements. Thus, an approach in robots/drones based on compound eye vision could be useful. In this research, several image processing algorithms were used to detect and track moving objects starting with images upon which a grid (compound eye image) was superimposed, including contours detection, the second moments of those contours along with the grid applied to the original image, and Fourier Transforms and inverse Fourier Transforms. The latter also provide information about scene or camera …

Automatic Gaze Classification For Aviators: Using Multi-Task Convolutional Networks As A Proxy For Flight Instructor Observation, Justin Wilson, Sandro Scielzo, Sukumaran Nair, Eric C. Larson

International Journal of Aviation, Aeronautics, and Aerospace

In this work, we investigate how flight instructors observe aviator scan patterns and assign quality to an aviator's gaze. We first establish the reliability of instructors to assign similar quality to an aviator's scan patterns, and then investigate methods to automate this quality using machine learning. In particular, we focus on the classification of gaze for aviators in a mixed-reality flight simulation. We create and evaluate two machine learning models for classifying gaze quality of aviators: a task-agnostic model and a multi-task model. Both models use deep convolutional neural networks to classify the quality of pilot gaze patterns for 40 …

A Mathematical Analysis Of The Game Of Santorini, Carson Clyde Geissler

Senior Independent Study Theses

Santorini is a two player combinatorial board game. Santorini bears resemblance to the graph theory game of Geography, a game of moving and deleting vertices on a graph. We explore Santorini with game theory, complexity theory, and artificial intelligence. We present David Lichtenstein’s proof that Geography is PSPACE-hard and adapt the proof for generalized forms of Santorini. Last, we discuss the development of an AI built for a software implementation of Santorini and present a number of improvements to that AI.

Cheat Detection Using Machine Learning Within Counter-Strike: Global Offensive, Harry Dunham

Senior Independent Study Theses

Deep learning is becoming a steadfast means of solving complex problems that do not have a single concrete or simple solution. One complex problem that fits this description and that has also begun to appear at the forefront of society is cheating, specifically within video games. Therefore, this paper presents a means of developing a deep learning framework that successfully identifies cheaters within the video game CounterStrike: Global Offensive. This approach yields predictive accuracy metrics that range between 80-90% depending on the exact neural network architecture that is employed. This approach is easily scalable and applicable to all types of …