Anomaly detection is the process of finding unusual events in a given dataset. Anomaly detection is often performed on datasets with a fixed set of predefined features. As a result of this, if the normal features bear a close resemblance to the anomalous features, most anomaly detection algorithms exhibit poor performance. This work seeks to answer the question, can we deform these features so as to make the anomalies standout and hence improve the anomaly detection outcome? We employ a Deep Learning and an Active Learning framework to learn features for anomaly detection. In Active Learning, an Oracle (usually a domain expert) labels a small amount of data over a series of training rounds. The deep neural network is trained after each round to incorporate the feedback from the Oracle into the model. Results on the MNIST, CIFAR-10 and Galaxy Zoo datasets show that our algorithm, Ahunt, significantly outperforms other anomaly detection algorithms used on a fixed, static, set of features. Ahunt can therefore overcome a poor choice of features that happen to be suboptimal for detecting anomalies in the data, learning more appropriate features. We also explore the role of the loss function and Active Learning query strategy, showing these are important, especially when there is a significant variation in the anomalies.

In this paper, we will introduce partial results from our 3-year NSF funded grant titled “Inverted and Active Learning Pedagogies (IALP) for Student Success.” We will present our results comparing student achievement between inverted (flipped) classrooms and traditional lecture formats in statistics classes at Texas Lutheran University. Included are a brief introduction, the study design, data gathering, faculty and students’ surveys, and the methodology used for the study. We will share results from instruments that were developed to measure different levels of cognitive understanding across multiple sections of the same introductory statistics classes. We will also examine grades, withdrawals, potential professor effects, student characteristics, class size effects, video saturation, accountability of students for watching videos, and student evaluations of various activities in the classroom.

Active participation in the learning process enhance students’ critical thinking and problem-solving skills. We implemented peerocused, active learning, recitation sessions with the large-enrollment sections for General Chemistry I courses at Mississippi State University (MSState) over a period of four semesters beginning in Spring 2016. The peerocused recitation program was a success improving student final (standardized ACS) exam scores, pass/fail rates for the course, and continuation on to General Chemistry II (CH 1223) courses. Peerocused collaborative learning and students possessing ownership over their learning significantly enhanced academic outcomes of our program. Worked-example effect is the best known and apparently the most effective cognitive load reducing technique. We incorporated a modified version of worked examples, employing “incorrect worked examples” and studied the impact of incorrect worked examples vs correct worked examples. We hypothesized that looking for errors in incorrect worked examples would achieve greater attention and would prompt students to actively engage on calculation steps than correct worked examples. Eye-tracking results showed that incorrect worked example format was effective at obtaining student attention and engaging students actively on calculation steps. Survey results showed that incorrect worked example format inspired students’ motivation and enhanced student engagement and attentiveness to examine the worked examples intensively. This research provided insights on student focus while reading and learning chemistry worked examples, and opened new avenues for supporting online learning and usage of tablet PC in the learning process. Laboratory experiments provide students the opportunity to obtain hands-on experience on laboratory techniques and instrumentation. We created a biochemistry laboratory course (CH4990) for third-year chemistry major undergraduate students at Mississippi State University. I wrote the biochemistry lab manual consisting of eleven experiments, which involved protein and DNA extraction, ion-exchange chromatography, UV/vis spectroscopy, SDS PAGE electrophoresis, and enzyme kinetics experiments. A new laboratory experiment was incorporated which allowed students exposure to peptide sequencing and proteomics experiments in conjunction with mass spectrometry. The CH4990 biochemistry lab course is open for enrollment in Fall semesters since Fall 2018.

No description available.

Location-based services have experienced substantial growth in the last decade. However, despite extensive research efforts, sub-meter location accuracy with low-cost infrastructure continues to be elusive. Comparing with infrastructure-based solutions, infrastructure-free indoor localization has the major advantage of avoiding extra cost for infrastructure deployment. There are two typical types of infrastructure-free indoor localization solutions, i.e., Pedestrian Dead Reckoning (PDR)-based and fingerprint-based. PDR-based solutions rely on inertial measurement units to estimate the user's relative location. Despite the effort, many issues still remain in PDR systems. For example, any deployed smartphone-based PDR system needs to cope with the changing orientation of smartphone that the phone might be putting in a pocket, or being taken out to use, etc. In addition, the outputs of Micro Electro-Mechanical Systems (MEMS) sensors on smart devices vary over time which results in rapidly accumulated localization errors without external references. Fingerprint-based solutions utilize different types of location dependent parameters to estimate user's absolute location. Although fingerprint-based solutions are usually more practical than PDR-based, they suffer from laborious site survey process. In this dissertation, we aim to mitigate these challenges. First of all, illumination intensity is introduced as a new type of fingerprints to provide location references for PDR-based indoor localization. We propose IDyLL -- an indoor localization system using inertial measurement units (IMU) and photodiode sensors on smartphones. Using a novel illumination peak detection algorithm, IDyLL augments IMU-based pedestrian dead reckoning with location fixes. Moreover, we devise a burned-out detection algorithm for simultaneous luminary-assisted IPS and burned-out luminary detection. Experimental study using data collected from smartphones shows that IDyLL is able to achieve high localization accuracy at low costs. As for fingerprint collection, several frameworks are proposed to ease the laborious site survey process, without compromising fingerprint quality. We propose TuRF, a path-based fingerprint collection mechanism for site survey. MobiBee, a treasure hunt game, is further designed to take advantage of gamification and incentive models for fast fingerprint collection. Motivated by applying mobile crowdsensing for fingerprint collection, we propose ALSense, a distributed active learning framework, for budgeted mobile crowdsensing applications. Novel stream-based active learning strategies are developed to orchestrate queries of annotation data and the upload of unlabeled data from mobile devices. Extensive experiments demonstrate that ALSense can indeed achieve higher classification accuracy given fixed data acquisition budgets. Facing malicious behaviors, three types of location-related attacks and their corresponding detection algorithms are investigated. Experiments on both crowdsensed and emulated dataset show that the proposed algorithms can detect all three types of attacks with high accuracy. / Thesis / Doctor of Philosophy (PhD)

Active learning is a practical field of machine learning as labeling data or determining which data to label can be a time consuming and inefficient task. Active learning automates the process of selecting which data to label, but current methods are heavily model reliant. This has led to the inability of sampled data to be transferred to new models as well as issues with sampling bias. Both issues are of crucial concern in machine learning deployment. We propose active learning methods utilizing Combinatorial Coverage to overcome these issues. The proposed methods are data-centric, and through our experiments we show that the inclusion of coverage in active learning leads to sampling data that tends to be the best in transferring to different models and has a competitive sampling bias compared to benchmark methods. / Master of Science / Machine learning (ML) models are being used frequently in a variety of applications. For the model to be able to learn, data is required. Processing this data is often one of the most, if not the most, time consuming aspects of utilizing ML. One especially burdensome aspect of data processing is data labeling, or determining what each data point corresponds to in terms of real world class. For example, determining if a data point that is an image contains a plane or bird. This way the ML model can learn from the data. Active learning is a sub-field of machine learning which aims to ease this burden by allowing the model to select which data would be most beneficial to label, so that the entirety of the dataset does not need to be labeled. The issue with current active learning methods is that they are highly model dependent. In machine learning deployment the model being used may change while data stays the same, so this model dependency can cause for data points we label with respect to one model to not be ideal for another model. This model dependency has led to sampling bias issues as well; points which are chosen to be labeled may all be similar or not representative of all data resulting in the ML model not being as knowledgeable as possible. Relevant work has focused on the sampling bias issue, and several methods have been proposed to combat this issue. Few of the methods are applicable to any type of ML model though. The issue of sampled points not generalizing to different models has been studied but no solutions have been proposed. In this work we present active learning methods using Combinatorial Coverage. Combinatorial Coverage is a statistical technique from the field of Design of Experiments, and has commonly been used to design test sets. The extension of Combinatorial Coverage to ML is newer, and provides a way to focus on the data. We show that this data focused approach to active learning achieves a better performance when the sampled data is used for a different model and that it achieves a competitive sampling bias compared to benchmark methods.

Intrusion detection systems are an integral part of web application security. As Internet use continues to increase, the demand for fast, accurate intrusion detection systems has grown. Various IDSs like Snort, Zeek, Solarwinds SEM, and Sleuth9, detect malicious intent based on existing patterns of attack. While these systems are widely deployed, there are limitations with their approach, and anomaly-based IDSs that classify baseline behavior and trigger on deviations were developed to address their shortcomings. Existing anomaly-based IDSs have limitations that are typical of any machine learning system, including high false-positive rates, a lack of clear infrastructure for deployment, the requirement for data to be centralized, and an inability to add modules tailored to specific organizational threats. To address these shortcomings, our work proposes a system that is distributed in nature, can actively learn and uses experts to improve accuracy. Our results indicate that the integrated system can operate independently as a holistic system while maintaining an accuracy of 99.03%, a false positive rate of 0.5%, and speed of processing 160,000 packets per second for an average system. / Master of Science / Intrusion detection systems are an integral part of web application security. The task of an intrusion detection system is to identify attacks on web applications. As Internet use continues to increase, the demand for fast, accurate intrusion detection systems has grown. Various IDSs like Snort, Zeek, Solarwinds SEM, and Sleuth9, detect malicious intent based on existing attack patterns. While these systems are widely deployed, there are limitations with their approach, and anomaly-based IDSs that learn a system's baseline behavior and trigger on deviations were developed to address their shortcomings. Existing anomaly-based IDSs have limitations that are typical of any machine learning system, including high false-positive rates, a lack of clear infrastructure for deployment, the requirement for data to be centralized, and an inability to add modules tailored to specific organizational threats. To address these shortcomings, our work proposes a system that is distributed in nature, can actively learn and uses experts to improve accuracy. Our results indicate that the integrated system can operate independently as a holistic system while maintaining an accuracy of 99.03%, a false positive rate of 0.5%, and speed of processing 160,000 packets per second for an average system.

Electronic textbooks are different from ebooks (electronic books) in that they allow the user to go beyond just reading material on a computer screen. Electronic textbooks encourage the user to accomplish all of the operations typically performed with a hardcopy text in addition to some functions not possible with paper books. With electronic textbooks users can make annotations in the textbook with e-ink; mark important sections; search over the ink, the text, or even the scanned images; look up items in online dictionaries or encyclopedias; perform interactive reinforcement drills; view simulations; and many other operations afforded by the computing power of the underlying computer and the reach of the Internet connection. These operations encourage students to engage in active reading. The VText framework is designed to provide many of the desired features of an e-textbook in such a way that it produces pedagogical value rather than just convenience for students. Many so-called e-textbook solutions available today provide few features beyond those possible with hardcopy textbooks. The VText framework is built as an add-in to Microsoft's note-taking program, OneNote. The add-in provides features that facilitate the use of OneNote as a reader and as an educational tool while leaving in place OneNote's strengths in note-taking, collaboration, and search. / Master of Science

Annotating large unlabeled datasets can be a major bottleneck for machine learning applications. We introduce a scheme for inferring labels of unlabeled data at a fraction of the cost of labeling the entire dataset. We refer to the scheme as Bounded Expectation of Label Assignment (BELA). BELA greedily queries an oracle (or human labeler) and partitions a dataset to find data subsets that have mostly the same label. BELA can then infer labels by majority vote of the known labels in each subset. BELA makes the decision to split or label from a subset by maximizing a lower bound on the expected number of correctly labeled examples. BELA improves upon existing hierarchical labeling schemes by using supervised models to partition the data, therefore avoiding reliance on unsupervised clustering methods that may not accurately group data by label. We design BELA with strategies to avoid bias that could be introduced through this adaptive partitioning. We evaluate BELA on labeling of four datasets and find that it outperforms existing strategies for adaptive labeling. / Master of Science / Most machine learning classifiers require data with both features and labels. The features of the data may be the pixel values for an image, the words in a text sample, the audio of a voice clip, and more. The labels of a dataset define the data. They place the data into one of several categories, such as determining whether a image is of a cat or dog, or adding subtitles to Youtube videos. The labeling of a dataset can be expensive, and usually requires a human to annotate. Human labeled data can be moreso expensive if the data requires an expert labeler, as in the labeling of medical images, or when labeling data is particularly time consuming. We introduce a scheme for labeling data that aims to lessen the cost of human labeled data by labeling a subset of an entire dataset and making an educated guess on the labels of the remaining unlabeled data. The labeled data generated from our approach may be then used towards the training of a classifier, or an algorithm that maps the features of data to a guessed label. This is based off of the intuition that data with similar features will also have similar labels. Our approach uses a game-like process of, at any point, choosing between one of two possible actions: we may either label a new data point, thus learning more about the dataset, or we may split apart the dataset into multiple subsets of data. We will eventually guess the labels of the unlabeled data by assigning each unlabeled data point the majority label of the data subset that it belongs to. The novelty in our approach is that we use supervised classifiers, or splitting techniques that use both the features and the labels of data, to split a dataset into new subsets. We use bias reduction techniques that enable us to use supervised splitting.

Using methodologies of interviews and observation, this study focuses on three distinctive and successful kinds of experiential education: a summer arts program, an outdoor science program and a wilderness education program. The project applies insights from the programs to the central question of this thesis: How can experiential learning be utilized within the traditional classroom given the constraints of the No Child Left Behind Law and standardized teaching?