Theory and Algorithms Commons^™

Improving Multi-Hop Knowledge Base Question Answering By Learning Intermediate Supervision Signals, Gaole He, Yunshi Lan, Jing Jiang, Wayne Xin Zhao, Ji Rong Wen

Research Collection School Of Computing and Information Systems

Multi-hop Knowledge Base Question Answering (KBQA) aims to find the answer entities that are multiple hops away in the Knowledge Base (KB) from the entities in the question. A major challenge is the lack of supervision signals at intermediate steps. Therefore, multi-hop KBQA algorithms can only receive the feedback from the final answer, which makes the learning unstable or ineffective. To address this challenge, we propose a novel teacher-student approach for the multi-hop KBQA task. In our approach, the student network aims to find the correct answer to the query, while the teacher network tries to learn intermediate supervision signals …

Optimizing Large-Scale Hyperparameters Via Automated Learning Algorithm, Bin Gu, Guodong Liu, Yanfu Zhang, Xiang Geng, Heng Huang

Machine Learning Faculty Publications

Modern machine learning algorithms usually involve tuning multiple (from one to thousands) hyperparameters which play a pivotal role in terms of model generalizability. Black-box optimization and gradient-based algorithms are two dominant approaches to hyperparameter optimization while they have totally distinct advantages. How to design a new hyperparameter optimization technique inheriting all benefits from both approaches is still an open problem. To address this challenging problem, in this paper, we propose a new hyperparameter optimization method with zeroth-order hyper-gradients (HOZOG). Specifically, we first exactly formulate hyperparameter optimization as an A-based constrained optimization problem, where A is a black-box optimization algorithm (such …

Unsupervised Data Mining Technique For Clustering Library In Indonesia, Robbi Rahim, Joseph Teguh Santoso, Sri Jumini, Gita Widi Bhawika, Daniel Susilo, Danny Wibowo

Library Philosophy and Practice (e-journal)

Organizing school libraries not only keeps library materials, but helps students and teachers in completing tasks in the teaching process so that national development goals are in order to improve community welfare by producing quality and competitive human resources. The purpose of this study is to analyze the Unsupervised Learning technique in conducting cluster mapping of the number of libraries at education levels in Indonesia. The data source was obtained from the Ministry of Education and Culture which was processed by the Central Statistics Agency (abbreviated as BPS) with url: bps.go.id/. The data consisted of 34 records where the attribute …

Modeling And Analysis Of Affiliation Networks With Subsumption, Alexey Nikolaev

Dissertations, Theses, and Capstone Projects

An affiliation (or two-mode) network is an abstraction commonly used for representing systems with group interactions. It consists of a set of nodes and a set of their groupings called affiliations. We introduce the notion of affiliation network with subsumption, in which no affiliation can be a subset of another. A network with this property can be modeled by an abstract simplicial complex whose facets are the affiliations of the network.

We introduce a new model for generating affiliation networks with and without subsumption (represented as simplicial complexes and hypergraphs, respectively). In this model, at each iteration, a constant number …

Divide And Capture: An Improved Cryptanalysis Of The Encryption Standard Algorithm Rsa, Willy Susilo, Joseph Tonien, Guomin Yang

Research Collection School Of Computing and Information Systems

RSA is a well known standard algorithm used by modern computers to encrypt and decrypt messages. In some applications, to save the decryption time, it is desirable to have a short secret key d compared to the modulus N. The first significant attack that breaks RSA with short secret key given by Wiener in 1990 is based on the continued fraction technique and it works with d < 1/4 root 18 N-.(25). A decade later, in 2000, Boneh and Durfee presented an improved attack based on lattice technique which works with d < N-.(292). Until this day, Boneh-Durfee attack remain as the best attack on RSA with short secret key. In this paper, we revisit the continued fraction technique and propose a new attack on RSA. Our main result shows that when d < root t (2 root 2 + 8/3) N-.(75)/root e, where e is the public exponent and t is a chosen parameter, our attack can break the RSA with the running time of O(tlog (N)). Our attack is especially well suited for the case where e is much smaller than N. When e approximate to N, the Boneh-Durfee attack outperforms ours. As a result, we could simultaneously run both attacks, our new attack and the classical Boneh-Durfee attack as a backup.

To Thine Own Self Be True? Incentive Problems In Personalized Law, Jordan M. Barry, John William Hatfield, Scott Duke Kominers

William & Mary Law Review

Recent years have seen an explosion of scholarship on “personalized law.” Commentators foresee a world in which regulators armed with big data and machine learning techniques determine the optimal legal rule for every regulated party, then instantaneously disseminate their decisions via smartphones and other “smart” devices. They envision a legal utopia in which every fact pattern is assigned society’s preferred legal treatment in real time.

But regulation is a dynamic process; regulated parties react to law. They change their behavior to pursue their preferred outcomes— which often diverge from society’s—and they will continue to do so under personalized law: They …

Norm-Based Generalisation Bounds For Deep Multi-Class Convolutional Neural Networks, Antoine Ledent, Waleed Mustafa, Yunwen Lei, Marius Kloft

Research Collection School Of Computing and Information Systems

We show generalisation error bounds for deep learning with two main improvements over the state of the art. (1) Our bounds have no explicit dependence on the number of classes except for logarithmic factors. This holds even when formulating the bounds in terms of the Frobenius-norm of the weight matrices, where previous bounds exhibit at least a squareroot dependence on the number of classes. (2) We adapt the classic Rademacher analysis of DNNs to incorporate weight sharing—a task of fundamental theoretical importance which was previously attempted only under very restrictive assumptions. In our results, each convolutional filter contributes only once …

Mimoa: A Membrane-Inspired Multi-Objective Algorithm For Green Vehicle Routing Problem With Stochastic Demands, Yunyun Niu, Yongpeng Zhang, Zhiguang Cao, Kaizhou Gao, Jianhua Xiao, Wen Song, Fangwei Zhang

Research Collection School Of Computing and Information Systems

Nowadays, an increasing number of vehicle routing problem with stochastic demands (VRPSD) models have been studied to meet realistic needs in the field of logistics. In this paper, a bi-objective vehicle routing problem with stochastic demands (BO-VRPSD) was investigated, which aims to minimize total cost and customer dissatisfaction. Different from traditional vehicle routing problem (VRP) models, both the uncertainty in customer demands and the nature of multiple objectives make the problem more challenging. To cope with BO-VRPSD, a membrane-inspired multi-objective algorithm (MIMOA) was proposed, which is characterized by a parallel distributed framework with two operation subsystems and one control subsystem, …

Fine-Grained Generalization Analysis Of Vector-Valued Learning, Liang Wu, Antoine Ledent, Yunwen Lei, Marius Kloft

Research Collection School Of Computing and Information Systems

Many fundamental machine learning tasks can be formulated as a problem of learning with vector-valued functions, where we learn multiple scalar-valued functions together. Although there is some generalization analysis on different specific algorithms under the empirical risk minimization principle, a unifying analysis of vector-valued learning under a regularization framework is still lacking. In this paper, we initiate the generalization analysis of regularized vector-valued learning algorithms by presenting bounds with a mild dependency on the output dimension and a fast rate on the sample size. Our discussions relax the existing assumptions on the restrictive constraint of hypothesis spaces, smoothness of loss …

Visual Analysis Of Discrimination In Machine Learning, Qianwen Wang, Zhenghua Xu, Zhutian Chen, Yong Wang, Shixia Liu, Huamin Qu

Research Collection School Of Computing and Information Systems

The growing use of automated decision-making in critical applications, such as crime prediction and college admission, has raised questions about fairness in machine learning. How can we decide whether different treatments are reasonable or discriminatory? In this paper, we investigate discrimination in machine learning from a visual analytics perspective and propose an interactive visualization tool, DiscriLens, to support a more comprehensive analysis. To reveal detailed information on algorithmic discrimination, DiscriLens identifies a collection of potentially discriminatory itemsets based on causal modeling and classification rules mining. By combining an extended Euler diagram with a matrix-based visualization, we develop a novel set …

Using Torchattacks To Improve The Robustness Of Models With Adversarial Training, William S. Matos Díaz

Cybersecurity: Deep Learning Driven Cybersecurity Research in a Multidisciplinary Environment

Adversarial training has proven to be one of the most successful ways to defend models against adversarial examples. This process consists of training a model with an adversarial example to improve the robustness of the model. In this experiment, Torchattacks, a Pytorch library made for importing adversarial examples more easily, was used to determine which attack was the strongest. Later on, the strongest attack was used to train the model and make it more robust against adversarial examples. The datasets used to perform the experiments were MNIST and CIFAR-10. Both datasets were put to the test using PGD, FGSM, and …

Hybrid Models As Transdisciplinary Research Enablers, Andreas Tolk, Alison Harper, Navonil Mustafee

Computational Modeling & Simulation Engineering Faculty Publications

Modelling and simulation (M&S) techniques are frequently used in Operations Research (OR) to aid decision-making. With growing complexity of systems to be modelled, an increasing number of studies now apply multiple M&S techniques or hybrid simulation (HS) to represent the underlying system of interest. A parallel but related theme of research is extending the HS approach to include the development of hybrid models (HM). HM extends the M&S discipline by combining theories, methods and tools from across disciplines and applying multidisciplinary, interdisciplinary and transdisciplinary solutions to practice. In the broader OR literature, there are numerous examples of cross-disciplinary approaches in …

The Complexity Of Symmetry, Matthew Lemay

HMC Senior Theses

One of the main goals of theoretical computer science is to prove limits on how efficiently certain Boolean functions can be computed. The study of the algebraic complexity of polynomials provides an indirect approach to exploring these questions, which may prove fruitful since much is known about polynomials already from the field of algebra. This paper explores current research in establishing lower bounds on invariant rings and polynomial families. It explains the construction of an invariant ring for whom a succinct encoding would imply that NP is in P/poly. It then states a theorem about the circuit complexity partial …

Going Meta On The Minimum Circuit Size Problem: How Hard Is It To Show How Hard Showing Hardness Is?, Zoë Bell

HMC Senior Theses

The Minimum Circuit Size Problem (MCSP) is a problem with a long history in computational complexity theory which has recently experienced a resurgence in attention. MCSP takes as input the description of a Boolean function f as a truth table as well as a size parameter s, and outputs whether there is a circuit that computes f of size ≤ s. It is of great interest whether MCSP is NP-complete, but there have been shown to be many technical obstacles to proving that it is. Most of these results come in the following form: If MCSP is NP-complete …

A Hybrid Gene Selection Strategy Based On Fisher And Ant Colony Optimization Algorithm For Breast Cancer Classification, Mohammed Hamim, Ismail El Moudden, Mohan D. Pant, Hicham Moutachaouik, Mustapha Hain

EVMS School of Health Professions Faculty Publications

Breast cancer poses the greatest threat to human life and especially to women's life. Despite the progress made in data mining technology in recent years, the ability to predict and diagnose such fatal diseases based on gene expression data still reveals a limited prediction performance, which may not be surprising since most of the genes in expression data are believed to be irrelevant or redundant. The dimensionality reduction process may be considered as a crucial step to analyze gene expression data, as it can reduce the high dimensionality of the breast cancer datasets, which may result into a better prediction …

A Literature Review Of Quantum Education In K-12 Level, Yuming He, Shenghua Zha, Wu He, Theo Bastiaens (Ed.)

Information Technology & Decision Sciences Faculty Publications

Quantum computing is an emerging technology paradigm of computing and has the potential to solve computational problems intractable using today’s classical computers or digital technology. Quantum computing is expected to be disruptive for many industries. The power of quantum computing technologies is based on the fundamentals of quantum mechanics, such as quantum superposition, quantum entanglement, or the no-cloning theorem. To build a highly trained and skilled quantum workforce that meets future industry needs, there is a need to introduce quantum concepts early on in K-12 schools since the learning of quantum is a lengthy process. As fundamental quantum concepts derive …

A Novel Dimensionality Reduction Approach To Improve Microarray Data Classification, Mohammed Hasim, Ismail El Mouden, Mounir Ouzir, Hicham Moutachaouik, Mustapha Hain

Department of Medicine Faculty Publications

Cancer tumor prediction and diagnosis at an early stage has become a necessity in cancer research, as it provides an increase in the treatment success chances. Recently, DNA microarray technology became a powerful tool for cancer identification, that can analyze the expression level of a different and huge number of genes simultaneously. In microarray data, the large genes number versus a few records may affect the prediction performance. In order to handle this "curse of dimensionality” constraint of microarray dataset while improving the cancer identification performance, a dimensional reduction phase is necessary. In this paper, we proposed a framework that …

Collateral Data Quality Challenges Of Iot Sensor-Generated Data, Richard Allen Herrin

Graduate Student Publications

Thousands of academic articles have been written about the various facets of the Internet of Things (IoT). Added to those are books of multiple flavors, conference proceedings, and a host of web-based content authored by a diverse cast of IoT community constituents. While there are many examples of successful IoT application solutions, participating technologies and how best to use them, are still relatively immature. These solutions are complex, geographically diverse, incorporate a broad spectrum of ever-evolving technologies that allow organizations to gather new data, create new value and do new things they haven’t been able to do effectively before.

Despite …

Xtreme-Noc: Extreme Gradient Boosting Based Latency Model For Network-On-Chip Architectures, Ilma Sheriff

All Graduate Theses, Dissertations, and Other Capstone Projects

Multiprocessor System-on-Chip (MPSoC) integrating heterogeneous processing elements (CPU, GPU, Accelerators, memory, I/O modules ,etc.) are the de-facto design choice to meet the ever-increasing performance/Watt requirements from modern computing machines. Although at consumer level the number of processing elements (PE) are limited to 8-16, for high end servers, the number of PEs can scale up to hundreds. A Network-on-Chip (NoC) is a microscale network that facilitates the packetized communication among the PEs in such complex computational systems. Due to the heterogeneous integration of the cores, execution of diverse (serial and parallel) applications on the PEs, application mapping strategies, and many other …

Novel Hedonic Games And Stability Notions, Jacob Schlueter

Theses and Dissertations--Computer Science

We present here work on matching problems, namely hedonic games, also known as coalition formation games. We introduce two classes of hedonic games, Super Altruistic Hedonic Games (SAHGs) and Anchored Team Formation Games (ATFGs), and investigate the computational complexity of finding optimal partitions of agents into coalitions, or finding - or determining the existence of - stable coalition structures. We introduce a new stability notion for hedonic games and examine its relation to core and Nash stability for several classes of hedonic games.