Physical Sciences and Mathematics Commons^™

Neural Net Stock Trend Predictor, Sonal Kabra

Master's Projects

This report analyzes new and existing stock market prediction techniques. Traditional technical analysis was combined with various machine-learning approaches such as artificial neural networks, k-nearest neighbors, and decision trees. Experiments we conducted show that technical analysis together with machine learning can be used to profitably direct an investor’s trading decisions. We are measuring the profitability of experiments by calculating the percentage weekly return for each stock entity under study. Our algorithms and simulations are developed using Python. The technical analysis methodology combined with machine learning algorithms show promising results which we discuss in this report.

An Open Source Discussion Group Recommendation System, Sarika Padmashali

Master's Projects

A recommendation system analyzes user behavior on a website to make suggestions about what a user should do in the future on the website. It basically tries to predict the “rating” or “preference” a user would have for an action. Yioop is an open source search engine, wiki system, and user discussion group system managed by Dr. Christopher Pollett at SJSU. In this project, we have developed a recommendation system for Yioop where users are given suggestions about the threads and groups they could join based on their user history. We have used collaborative filtering techniques to make recommendations and …

Credit Scoring Using Logistic Regression, Ansen Mathew

Master's Projects

This report presents an approach to predict the credit scores of customers using the Logistic Regression machine learning algorithm. The research objective of this project is to perform a comparative study between feature selection and feature extraction, against the same dataset using the Logistic Regression machine learning algorithm. For feature selection, we have used Stepwise Logistic Regression. For feature extraction, we have used Singular Value Decomposition (SVD) and Weighted Singular Value Decomposition (SVD). In order to test the accuracy obtained using feature selection and feature extraction, we used a public credit dataset having 11 features and 150,000 records. After performing …

Document Classification Using Machine Learning, Ankit Basarkar

Master's Projects

To perform document classification algorithmically, documents need to be represented such that it is understandable to the machine learning classifier. The report discusses the different types of feature vectors through which document can be represented and later classified. The project aims at comparing the Binary, Count and TfIdf feature vectors and their impact on document classification. To test how well each of the three mentioned feature vectors perform, we used the 20-newsgroup dataset and converted the documents to all the three feature vectors. For each feature vector representation, we trained the Naïve Bayes classifier and then tested the generated classifier …

Ai For Classic Video Games Using Reinforcement Learning, Shivika Sodhi

Master's Projects

Deep reinforcement learning is a technique to teach machines tasks based on trial and error experiences in the way humans learn. In this paper, some preliminary research is done to understand how reinforcement learning and deep learning techniques can be combined to train an agent to play Archon, a classic video game. We compare two methods to estimate a Q function, the function used to compute the best action to take at each point in the game. In the first approach, we used a Q table to store the states and weights of the corresponding actions. In our experiments, this …

Comparing Authentic And Cryptic 5’ Splice Sites Using Hidden Markov Models And Decision Trees, Pratikshya Mishra

Master's Projects

Splicing is the editing of the precursor mRNA produced during transcription. The mRNA contains a large number of nucleotides in the introns and exons which are spliced to remove the introns and bind the exons to produce the mature mRNA which is translated to generate proteins. Hence accurate splicing at 5’ and 3’ splice sites (authentic splice sites (AuthSS)) is of foremost importance. The 5’ and 3’ splice sites are characterized by consensus sequences. Eukaryotic genome also contains splice sites known as Cryptic Splice Sites (CSS) that match the consensus. But the CSS are activated only when there is a …

Computational Analysis Of Cryptic Splice Sites, Remya Mohanan

Master's Projects

DNA in the nucleus of all eukaryotes is transcribed into mRNA where it is then translated into proteins. The DNA which is transcribed into mRNA is composed of coding and non-coding regions called exons and introns, respectively. It undergoes a post-trancriptional process called splicing where the introns or the non-coding regions are removed from the pre-mRNA to give the mature mRNA. Splicing of pre-mRNAs at 5 ́ and 3ˊ ends is a crucial step in the gene expression pathway. The mis-splicing by the spliceosome at different sites known as cryptic splice sites is caused by mutations which will affect the …

Generic Online Learning For Partial Visible & Dynamic Environment With Delayed Feedback, Behrooz Shahriari

Master's Projects

Reinforcement learning (RL) has been applied to robotics and many other domains which a system must learn in real-time and interact with a dynamic environment. In most studies the state- action space that is the key part of RL is predefined. Integration of RL with deep learning method has however taken a tremendous leap forward to solve novel challenging problems such as mastering a board game of Go. The surrounding environment to the agent may not be fully visible, the environment can change over time, and the feedbacks that agent receives for its actions can have a fluctuating delay. In …

Headline Generation Using Deep Neural Networks, Dhruven Vora

Master's Projects

News headline generation is one of the important text summarization tasks. Human generated news headlines are generally intended to catch the eye rather than provide useful information. There have been many approaches to generate meaningful headlines by either using neural networks or using linguistic features. In this report, we are proposing a novel approach based on integrating Hedge Trimmer, which is a grammar based extractive summarization system with a deep neural network abstractive summarization system to generate meaningful headlines. We analyze the results against current recurrent neural network based headline generation system.

Cascaded Facial Detection Algorithms To Improve Recognition, Edmund Yee

Master's Projects

The desire to be able to use computer programs to recognize certain biometric qualities of people have been desired by several different types of organizations. One of these qualities worked on and has achieved moderate success is facial detection and recognition. Being able to use computers to determine where and who a face is has generated several different algorithms to solve this problem with different benefits and drawbacks. At the backbone of each algorithm is the desire for it to be quick and accurate. By cascading face detection algorithms, accuracy can be improved but runtime will subsequently be increased. Neural …

Shopbot: An Image Based Search Application For E-Commerce Domain, Nishant Goel

Master's Projects

For the past few years, e-commerce has changed the way people buy and sell products. People use this business model to do business over the Internet. In this domain, Human-Computer Interaction has been gaining momentum. Lately, there has been an upsurge in agent based applications in the form of intelligent personal assistants (also known as Chatbots) which make it easier for users to interact with digital services via a conversation, in the same way we talk to humans. In e- commerce, these assistants offer mainly text-based or speech based search capabilities. They can handle search for most products, but cannot …

Application Of Computational Methods To Study The Selection Of Authentic And Cryptic Splice Sites, Tapomay Dey

Master's Projects

Proteins are building blocks of the bodies of eukaryotes, and the process of synthesizing proteins from DNA is crucial for the good health of an organism [13]. However, some mutations in the DNA may disrupt the selection of 5’ or 3’ splice sites by a spliceosome. An important research question is whether the disruptions have a stochastic relation to the position of nucleotides in the vicinity of the known authentic and cryptic splice sites. This can be achieved by proving that the authentic and cryptic splice sites are intrinsically different. However, the behavior of the spliceosome is not accurately known. …

A Chatbot Framework For Yioop, Harika Nukala

Master's Projects

Over the past few years, messaging applications have become more popular than Social networking sites. Instead of using a specific application or website to access some service, chatbots are created on messaging platforms to allow users to interact with companies’ products and also give assistance as needed. In this project, we designed and implemented a chatbot Framework for Yioop. The goal of the Chatbot Framework for Yioop project is to provide a platform for developers in Yioop to build and deploy chatbot applications. A chatbot is a web service that can converse with users using artificial intelligence in messaging platforms. …

Mining Frequency Of Drug Side Effects Over A Large Twitter Dataset Using Apache Spark, Dennis Hsu

Master's Projects

Despite clinical trials by pharmaceutical companies as well as current FDA reporting systems, there are still drug side effects that have not been caught. To find a larger sample of reports, a possible way is to mine online social media. With its current widespread use, social media such as Twitter has given rise to massive amounts of data, which can be used as reports for drug side effects. To process these large datasets, Apache Spark has become popular for fast, distributed batch processing. In this work, we have improved on previous pipelines in sentimental analysis-based mining, processing, and extracting tweets …

Masquerade Detection On Mobile Devices, Swathi Nambiar Kadala Manikoth

Master's Projects

A masquerade is an attack where the attacker avoids detection by impersonating an authorized user of a system. In this research we consider the problem of masquerade detection on mobile devices. Our goal is to improve on previous work by considering more features and a wide variety of machine learning techniques. Our approach consists of verifying the authenticity of users based on individual features and combinations of features for all users to determine which features contribute the most to masquerade detection. Also, we determine which of the two approaches - the combination of features or using individual features has performed …

Named Entity Recognition And Classification For Natural Language Inputs At Scale, Shreeraj Dabholkar

Master's Projects

Natural language processing (NLP) is a technique by which computers can analyze, understand, and derive meaning from human language. Phrases in a body of natural text that represent names, such as those of persons, organizations or locations are referred to as named entities. Identifying and categorizing these named entities is still a challenging task, research on which, has been carried out for many years. In this project, we build a supervised learning based classifier which can perform named entity recognition and classification (NERC) on input text and implement it as part of a chatbot application. The implementation is then scaled …

Image Spam Detection, Aneri Chavda

Master's Projects

Email is one of the most common forms of digital communication. Spam can be de ned as unsolicited bulk email, while image spam includes spam text embedded inside images. Image spam is used by spammers so as to evade text-based spam lters and hence it poses a threat to email based communication. In this research, we analyze image spam detection methods based on various combinations of image processing and machine learning techniques.

An Improved Algorithm For Learning To Perform Exception-Tolerant Abduction, Mengxue Zhang

McKelvey School of Engineering Theses & Dissertations

Abstract

Inference from an observed or hypothesized condition to a plausible cause or explanation for this condition is known as abduction. For many tasks, the acquisition of the necessary knowledge by machine learning has been widely found to be highly effective. However, the semantics of learned knowledge are weaker than the usual classical semantics, and this necessitates new formulations of many tasks. We focus on a recently introduced formulation of the abductive inference task that is thus adapted to the semantics of machine learning. A key problem is that we cannot expect that our causes or explanations will be perfect, …

Measuring Presence In A Police Use Of Force Simulation, Dharmesh Rajendra Desai

University of New Orleans Theses and Dissertations

We have designed a simulation that can be used to train police officers. Digital simulations are more cost-effective than a human role play. Use of force decisions are complex and made quickly, so there is a need for better training and innovative methods. Using this simulation, we are measuring the degree of presence that a human experience in a virtual environment. More presence implies better training. Participants are divided into two groups in which one group performs the experiment using a screen, keyboard, and mouse, and another uses virtual reality controls. In this experiment, we use subjective measurements and physiological …

Predicting User Choices In Interactive Narratives Using Indexter's Pairwise Event Salience Hypothesis, Rachelyn Farrell

University of New Orleans Theses and Dissertations

Indexter is a plan-based model of narrative that incorporates cognitive scientific theories about the salience—or prominence in memory—of narrative events. A pair of Indexter events can share up to five indices with one another: protagonist, time, space, causality, and intentionality. The pairwise event salience hypothesis states that when a past event shares one or more of these indices with the most recently narrated event, that past event is more salient, or easier to recall, than an event which shares none of them. In this study we demonstrate that we can predict user choices based on …

Large-Scale Discovery Of Visual Features For Object Recognition, Drew Linsley, Sven Eberhardt, Dan Shiebler, Thomas Serre

MODVIS Workshop

A central goal in vision science is to identify features that are important for object and scene recognition. Reverse correlation methods have been used to uncover features important for recognizing faces and other stimuli with low intra-class variability. However, these methods are less successful when applied to natural scenes with variability in their appearance.

To rectify this, we developed Clicktionary, a web-based game for identifying features for recognizing real-world objects. Pairs of participants play together in different roles to identify objects: A “teacher” reveals image regions diagnostic of the object’s category while a “student” tries to recognize the object. Aggregating …

Harnessing Predictive Models For Assisting Network Forensic Investigations Of Dns Tunnels, Irvin Homem, Panagiotis Papapetrou

Annual ADFSL Conference on Digital Forensics, Security and Law

In recent times, DNS tunneling techniques have been used for malicious purposes, however network security mechanisms struggle to detect them. Network forensic analysis has been proven effective, but is slow and effort intensive as Network Forensics Analysis Tools struggle to deal with undocumented or new network tunneling techniques. In this paper, we present a machine learning approach, based on feature subsets of network traffic evidence, to aid forensic analysis through automating the inference of protocols carried within DNS tunneling techniques. We explore four network protocols, namely, HTTP, HTTPS, FTP, and POP3. Three features are extracted from the DNS tunneled traffic: …

Search In T Cell And Robot Swarms: Balancing Extent And Intensity, George M. Fricke

Computer Science ETDs

This work investigates effective search and resource collection algorithms for swarms. Deterministic spiral algorithms and L ́evy search processes have been shown to be optimal for single searchers. We extend these strategies to swarms of robots and populations of T cells and measure performance under a variety of conditions.

Search extent and intensity lie on a continuum: more intensive patterns search thoroughly in the local area, while extensive patterns cover more area but may miss targets nearby. We show that the most efficient trade-off between search intensity and extent for swarms depends strongly on the distribution of targets, swarm size …

Bayesian Optimization For Refining Object Proposals, With An Application To Pedestrian Detection, Anthony D. Rhodes

Student Research Symposium

We devise an algorithm using a Bayesian optimization framework in conjunction with contextual visual data for the efficient localization of objects in still images. Recent research has demonstrated substantial progress in object localization and related tasks for computer vision. However, many current state-of-the-art object localization procedures still suffer from inaccuracy and inefficiency, in addition to failing to successfully leverage contextual data. We address these issues with the current research.

Our method encompasses an active search procedure that uses contextual data to generate initial bounding-box proposals for a target object. We train a convolutional neural network to approximate an offset distance …

Curiosity: Emergent Behavior Through Interacting Multi-Level Predictions, Douglas S. Blank, Lisa Meeden, James Marshall

Computer Science Faculty Research and Scholarship

Over the past 15 years our research group has been exploring models of developmental robotics and curiosity. Our research is based on the premise that intelligent behavior arises through emergent interactions between opposing forces in an open-ended, task-independent environment. In an initial experiment we constructed a recurrent neural network model where self-motivation was "an emergent property generated by the competing pressures that arise in attempting to balance predictability and novelty". The system first focused on its error, then learned to successfully predict its error, and finally became habituated to what caused the error. This process of focusing, learning, and habituating …

Comparing Tensorflow Deep Learning Performance Using Cpus, Gpus, Local Pcs And Cloud, John Lawrence, Jonas Malmsten, Andrey Rybka, Daniel A. Sabol, Ken Triplin

Publications and Research

Deep learning is a very computational intensive task. Traditionally GPUs have been used to speed-up computations by several orders of magnitude. TensorFlow is a deep learning framework designed to improve performance further by running on multiple nodes in a distributed system. While TensorFlow has only been available for a little over a year, it has quickly become the most popular open source machine learning project on GitHub. The open source version of TensorFlow was originally only capable of running on a single node while Google’s proprietary version only was capable of leveraging distributed systems. This has now changed. In this …

Investigating Trust And Trust Recovery In Human-Robot Interactions, Abigail L. Thomson

Celebration of Learning

As artificial intelligence and robotics continue to advance and be used in increasingly different functions and situations, it is important to look at how these new technologies will be used. An important factor in how a new resource will be used is how much it is trusted. This experiment was conducted to examine people’s trust in a robotic assistant when completing a task, how mistakes affect this trust, and if the levels of trust exhibited with a robot assistant were significantly different than if the assistant were human. The task was to watch a computer simulation of the three-cup monte …

Robot Learning From Human Demonstration: Interpretation, Adaptation, And Interaction, Chi Zhang

Doctoral Dissertations

Robot Learning from Demonstration (LfD) is a research area that focuses on how robots can learn new skills by observing how people perform various activities. As humans, we have a remarkable ability to imitate other human’s behaviors and adapt to new situations. Endowing robots with these critical capabilities is a significant but very challenging problem considering the complexity and variation of human activities in highly dynamic environments.

This research focuses on how robots can learn new skills by interpreting human activities, adapting the learned skills to new situations, and naturally interacting with humans. This dissertation begins with a discussion of …

Image Segmentation Using De-Textured Images, Yaswanth Kodavali

All Graduate Plan B and other Reports, Spring 1920 to Spring 2023

Image segmentation is one of the fundamental problems in computer vision. The outputs of segmentation are used to extract regions of interest and carry out identification or classification tasks. For these tasks to be reliable, segmentation has to be made more reliable. Although there are exceptionally well-built algorithms available today, they perform poorly in many instances by producing over-merged (combining many unrelated objects) or under-merged (one object appeared as many) results. This leads to far fewer or more segments than expected. Such problems primarily arise due to varying textures within a single object and/or common textures near borders of adjacent …

Music Feature Matching Using Computer Vision Algorithms, Mason Hollis

Computer Science and Computer Engineering Undergraduate Honors Theses

This paper seeks to establish the validity and potential benefits of using existing computer vision techniques on audio samples rather than traditional images in order to consistently and accurately identify a song of origin from a short audio clip of potentially noisy sound. To do this, the audio sample is first converted to a spectrogram image, which is used to generate SURF features. These features are compared against a database of features, which have been previously generated in a similar fashion, in order to find the best match. This algorithm has been implemented in a system that can run as …