Browsing by Subject "Online learning"
Now showing 1 - 20 of 23
- Results Per Page
- Sort Options
Item A critical consideration of contemporary openness in online education(2018-08) Dearmon, Matthew Jacob; De Lissovoy, Noah, 1968-; Pinar, William F; Brown, Anthony L; Hughes, Joan E; Salinas, CinthiaAs the so-called Open Education Movement develops amidst the maturation and growth of computer and Internet technologies, there exists a need for a critical understanding of Open Education itself and its implications for online learning and teaching at distance and scale. To that end, this project essays to establish the limits and possibilities of Open Education as they exist within the context of contemporary neoliberal ideological infiltration of public and higher education, as well as associated processes and structures of licensing, funding, and curriculum. Utilizing a deschooled critical approach grounded in postmodern theories of rhizomatic formation and contemporary notions of the commons, this textual and theoretical research begins by stating the need to clarify what is meant by the term, "open education" and verifying whether and to what extent existing scholarship has engaged the subject at a level appropriate to the threat posed by neoliberal policies, discourses, practices, and enclosures. Applying a transformative research paradigm to a textual analysis that views purposefully-selected free-and-open learning, education, and teaching websites as examples of material culture, this research project seeks to understand Open Education outside of the strictures and limitations of institutionalized education. By examining the mission of selected sites, their promotion of open licensing practices, funding resources that make such learning possible, curricular decisions made at networked scale, and a sample of learning experiences, a conception of limits and possibilities emerges within each of these domains. It is suggested that by encouraging reciprocal learning and teaching through the most permissive level of attributive licensing that encourages sharing, open education can indeed realize some measure of its potential to proliferate open and inclusive learning practices at scale. Due to its low barrier of entry, relative openness, and non-reliance on institutionalized funding, Wikiversity is promoted as a promising site for future efforts through a model of Openly Shared Learning Opportunities (OSLO), even though continued care must be taken to resist corporatocratic and neoliberal intrusion. By removing traditional boundaries established by the need for "teachers" to "educate" learners, OSLO reinvigorates both the subject and the Multitude through engagement with the digital commons.Item A fraction intervention using virtual manipulatives in synchronous online learning for students with mathematics learning disabilities(2021-08-25) Park, Jiyeon (Ph. D. in special education); Bryant, Diane Pedrotty; Shin, Mikyung; Falcomata, Terry; Doabler, ChristopherOnline learning provides customized instruction with high flexibility and accessibility, but research into the delivery of online interventions for students with disabilities remains scarce. The purpose of this study is to investigate the effects of fraction intervention that pair virtual manipulatives with instructional lessons featuring explicit instruction in synchronous online learning for fifth grade students with mathematics learning disabilities (MLD). A multiple probe baseline design across subjects was employed for the study. Three fifth-grade students with MLD received explicit instruction in synchronous online learning via video conferencing programs. During the intervention, students received the 15 fraction lessons using virtual manipulatives (VMs) in synchronous online learning environment. The participants’ percentages of correct answers on the researcher-developed probe were measured across baseline, intervention, and maintenance phases. After the intervention, the participants completed a social validity questionnaire which examined their perceptions of the online intervention using VMs. The findings of the study showed that students' performance improved when the intervention was introduced, and that there was a functional relationship between the fraction intervention and students’ accuracy in solving problems on equivalent fractions. Specifically, the average performance levels across all participants increased from the baseline to the intervention phase (by 18.33% to 20%) with high effect sizes (PND of 80%-100%, NAP of 0.94-1.00, and Tau-U of 0.88-1.00). However, the extent of the effects and maintenance varied according to students' participation and perspectives toward online learning.Item Adaptive learning approaches for smart home environments with a simulator implementation(2018-05-04) Dinc, Ahsen; Julien, Christine, D. Sc.Smart home solutions are utilized in a different way by each user according to the user’s unique needs and preferences over his unique home setting. User – device interactions themselves carry some implicit characteristics of the context the smart devices are used. In order to better exploit Internet of Things technologies in smart home environments, the user’s interaction history can be leveraged to generate knowledge of his behavior habits. With a purpose of achieving personalized decision making of smart lighting environments, this thesis presents three learning model approaches, which are based solely on the individual’s interaction habits with the devices, adaptive to changes in inhabitant’s device usage behavior, and do not require preset data for initialization. Individual interactions with smart light devices are contextualized based on the timestamp of the interaction, and ambient light intensity reading of the room at that instant. Moreover, the thesis introduces a Java simulator, which interactively demonstrates the behavior of a personalized smart home setting with the integrated learning models with a feedback mechanism. The simulator leads up to a way of evaluating adaptive learning models in smart home environments, reducing the immediate need of testing their behavior in a real life smart home, which is both hard and costly to construct and maintain. The learning approaches, namely K-Nearest Neighbor and Softmax Regression with batch learning and online learning algorithms, are evaluated with different datasets representing different scenarios, and the results show that all three methods are able to perfectly capture the usage pattern when the interactions with distinct “things”, smart light devices, are separable in terms of their corresponding context. For more complex datasets, which have overlaps between the usage context of distinct devices and big changes in user behavior over time, the online learning algorithm needs more data in order to catch the performance of KNN and Softmax Regression with batch learning algorithms.Item Building bridges to engineering interest : instructor inputs and situational interest in online engineering classes(2021-11-29) Miesner, Ella K.; Schallert, Diane L.; Chang, Seung W; French, Karen D; Yan, Veronica XDevelopment of subject matter interest is an important motivator for long term student learning. Fortunately, what teachers do in the learning environment can have an impact on interest. This study was an investigation of how instructor inputs in engineering courses that were taught online were associated with students’ perceptions of instructor relatedness and situational interest. The initial hypothesis was that a student’s sense of social relatedness with the instructor would mediate the relationship between instructional inputs and student perceptions of situational interest in the class. The study employed a multiple methods approach to gather quantitative and qualitative data to gain insights into both instructor and student perspectives of the online engineering educational experience. The frameworks of teacher presence, defined by Garrison’s (2007) Community of Inquiry (COI) model as the design and facilitation of distance learning environments, and Mehrabian's (1971) concept of immediacy, which includes behaviors intended to signal openness to communicative interaction, were used to develop a survey administered to 141 undergraduate engineering students, capturing their perceptions of social relatedness and situational interest. Results indicated that teacher presence online, but not immediacy behaviors, was predictive of student-perceived instructor relatedness and situational interest. A higher reported perception of teacher presence was associated with a greater feeling of instructor-student relatedness and with situational interest. However, the two outcome variables were independent such that instructor relatedness did not mediate the relationship between instructor inputs and interest. Follow-up qualitative interviews indicated that the teacher presence construct provided a sufficient framework to encompass student and instructor perceptions of the role of the instructor, as well as strategies used by instructors for developing instructor-student connection and the encouragement of situational interest during class sessions. A synthesis of the qualitative and quantitative data suggests that instructors can use specific strategies from the teacher presence framework, including a focus on instructional design, facilitation of discourse, and well-organized direct instruction to help students develop feelings of both connection and interest. Although the initial hypothesis of a mediated path to interest via the development of social relatedness was not supported, the final results suggest that both instructor-student relatedness and situational interest develop concurrently and that the affective and cognitive components of online engineering education should be considered simultaneously.Item Course design and student learning outcomes in asynchronous online courses : the role of student motivation(2022-05-02) Shi, Yi (Ph. D. in curriculum and instruction); Hughes, Joan E.; Liu, Min; Riegle-Crumb, Catherine; Muenks, Katherine MThe purpose of the study is to reveal instructors’ design strategies regarding dialogue and structure in asynchronous online courses, uncover how those designs were received by students, and the predictive effects of course design and student motivation on student learning. A sample of 287 college students and two online course instructors participated in the study. The findings showed that course formality and flexibility indirectly predicted student perceived learning outcomes through the roles of subject specific task values, subject specific emotional cost, online learning self-efficacy, and online learning task values. Dialogue with instructors and classmates did not significantly predict student perceived learning outcomes. Results of student and instructor interviews supported the quantitative results and revealed that formality and flexibility of asynchronous online courses were highly appreciated by students, whereas the dialogic practices instructors used received mixed feedback. These findings suggest that course design is critical in asynchronous online learning, and it has positive impact on students’ motivation, emotional well-being, and student learning outcomes. Practical implications for designing asynchronous online courses are discussed in the paper.Item Educator perceptions during implementation of blended learning : a high school case study(2018-03-26) Cooper, Beth Ann; Olivárez, Rubén; Sharpe, Edwin; Pringle, Patrick; Somers, PatriciaImplementation of blended learning in K-12 education advanced rapidly without significant research to guide leaders in decision making and planning. Studying blended learning through the Implementation Stages framework developed by the SISEP Center allows educators to evaluate best practices for implementing this innovative instructional strategy and to determine necessary components for scaling the model. This qualitative case study at a Texas public high school offers a view of educators’ perceptions during the implementation of The University of Texas at Austin High School’s digital curriculum in a blended learning pilot. Three research questions drive this study: (1) What are educators’ perceptions of blended learning during installation and initial implementation stages? (2) How do educators define student success during blended learning implementation stages? (3) In which areas can educators’ perceptions during installation and initial implementation of blended learning inform district decisions regarding full implementation? The study incorporates a qualitative methodology built from a constructivist approach, recognizing the value of individuals’ meaning making processes to form a collective picture. Three data sources consisted of surveys, interviews, and document reviews. An organized process to code information into themes generated areas for focus while answering each research question with rich, thick description. The findings showed that educators discussed four key themes when reflecting on the blended learning implementation including (a) training and support, (b) aligned goals; (c) educator access; and (d) personnel and mindset. They revealed four themes for ways of defining student success consisting of (a) mastery of core standards; (b) student access; (c) learner personalization; and (d) 21st century skills. Four integrated themes emerged to guide district leaders in decision-making processes to determine whether to move forward with full implementation of the pilot project. These areas for focus include (a) power of the people; (b) aligned integration; (c) personalization pursuits; and (d) blended bandwagon. Although the case study district chose not to scale the pilot project to full implementation, the case offers insight into the processes and areas for focus in decisionmaking and guiding future research in the area of implementation of innovative programs and blended learningItem The effects of online collaborative learning activities on student perception of level of engagement(2006-12) Quiros, Ondrea Michelle; Resta, Paul E.As online learning becomes more popular, higher education is becoming more interested in this new medium of learning. However, attrition has become a developing problem for colleges and universities that offer online classes, as some students found it was difficult to stay engaged in their online courses. From the literature, it was hypothesized that instructional designs that incorporate collaborative activities will lead to higher perceived engagement levels than those that incorporate individualistic learning. An exploratory study used a self-report survey instrument to measure students' perception of level of engagement in six graduate-level online classes (n=66). Half of the courses in the study integrated formal collaborative activities as a significant component of the course and half represented learning environments characterized by whole group and individualistic learning. The results showed a significant positive relationship between classes that used collaborative activities and engagement levels. However, the coded responses of the participants showed that while classes that use such activities had higher levels of engagement, it is possible that this may be attributable to other factors external to the formal elements of collaboration in the course. Recommendations are offered for future research that may help identify the elements that contribute to engagement in online courses.Item Extending the online distance course : online student activity beyond the online classroom(2012-12) Barrera, Rachel Edith; Hughes, Joan E.; Svinicki, Marilla; Field, Sherry; Resta, Paul; Lewis, KarronThis study investigated why and how students, who enrolled in fully-online distance course, participated in online activities external to the formal online course (OAEOC) at any point during or after the online course. For this research, OAEOC is defined as any activity pursued by students within an online environment during or after the course that does not take place within their teacher-sponsored online course “home” (such as a Moodle or Blackboard). This research occurred within a fully-online, five-week course that trained journalists in digital tools. Data included: (a) 144 researcher-generated interpretive memos based on activities within the course’s online discussion forums and student chats and (b) 11 student interviews. Results showed that student interactions in course discussion forums were critically important for developing connections between students, which in turn, supported the initiation of online activities external to the online course. During the course, students posted information about their online identities and created a Facebook group and Twitter list, which facilitated online activities external to the course. Data from interviews showed that those students participating in OAEOC did so for social reasons and to continue conversing with classmates. Students who did not participate in OAEOCs indicated work schedule conflicts, lack of interest, and unawareness of the OAEOCs prevented their participation. During the course, OAEOC participants discussed topics related to the course content. However, once the course concluded, OAEOC participants started discussing more personal and professional topics. The phenomenon studied is new to online distance education and holds the potential to extend the online course experience and support lifelong learning.Item Factors affecting faculty technology adoption of online teaching in higher education : literature review(2012-05) Zhou, Yonghan; Liu, Min, Ed. D.; Resta, PaulOnline teaching and learning has grown rapidly in current educational contexts. Whereas once, the role of faculty was primarily a classroom instructor, in online classrooms, the role has been expanded to one of facilitator, organizer, and supporter. The more efficiently that faculty can adopt online technology and apply it to their teaching and instruction, the better students academic results will achieve (Goktalay & Huguet, 2006). The purpose of this literature review is to help faculty members to adopt new online technologies more effectively and successfully. This literature review identifies important factors that contribute to faculty members’ adoption of technology in higher education. Among these factors are: reliability of online technology, faculty’s perceived usefulness of technology, institutional support of online technology, time constraints in implementing online technology to instructional methods and developing effective goals for the use of technology, and then provides recommendations based on these affecting factors.Item Meeting students' needs and expectations in a culturally diverse e-learning environment : a case study(2010-12) Pham, Minh Trung; Liu, Min, Ed. D.; Cummings, Wm. Theodore; Hughes, Joan; Northcutt, Norvell; Resta, Paul E.The increased growth of online instruction has been well documented by various studies. As the result of the proliferation of online instruction, students from outside of the United States are now able to obtain an American education without having to leave their home country. While online course designs have been well researched and documented to identify best methods and practices to enable optimum learning achievement, providing online instruction to non-US educated students generates the question of how a culturally diverse student body adapts and/or adjusts to an American-style instruction. The purpose of this study is to conduct an exploratory qualitative research to investigate how students from an Asian learning culture adapt to an American online learning environment and to determine whether the various instructional design theories and practices that are widely accepted as best practices in the United States and incorporated into the instruction designs for this Marketing Management hybrid course are also as well-received by students from a different learning culture. From the five categories emerged from the research data: (1) students’ background, (2) perceived benefits, (3) essential skills, (4) supports expected and/or received, and (5) sense of community, the researcher proposed a framework that encompasses the students’ process of adapting to online learning. Within the process of adapting to online learning, conditions such as students’ backgrounds and expectations influenced the various learning strategies that students adopted in order to realize the benefits from the online learning experience. Information gathered from this study may provide those involved in online education - decision makers in academic, business, and professional organizations considering an overseas online instruction strategy - an added awareness of how different learning cultures may influence the quality of an online learning experience. Additionally, for a specific target audience, this research study may further validate the learner-centered approach for instruction designs. For students who may be contemplating online learning as an option, this study may provide a deeper understanding of what is entailed in an online learning environment - the contributing actors and factors that affect the quality of an online learning experience.Item Minding the verge: moderating webcasts+chat in a multi-section online undergraduate course(2009-08) Hamerly, Donald Wade; Immroth, Barbara FrolingCoincidental increases in online instruction at institutions of higher education and in online social networking generally in the U.S. have created opportunities for research into how digital interpersonal connectivity affects online learning. This study examined interactive webcasts, or webcasts plus chat, that were part of an online undergraduate course covering Internet knowledge and skills at a large public university. Symbolic interactionism served as the theoretical framework for explicating interactive webcasts as useful online learning environments by exploring the complex processes that instructional staff employed to manage their actions and interactions as moderators in the webcasts and chats. A constructivist grounded theory approach guided the collection and analysis of empirical data in the form of webcast media and transcripts, chat logs, students‘ reflective writing, and semi-structured, intensive interviews with instructional staff. From the study emerged theoretical categories in three tiers related to a generalized moderator process called minding the verge: moderators minded the verge in three conditions of interaction– converging, attending, and diverging; in three loci of interaction – webcasts, chats, and webcasts+chat; and through six actions of moderating – bonding, orientating, guiding, tending, validating, and branching. The results of this study provide moderators for the course with insights into their actions in the interactive webcasts and with concepts moderators can use to explore how to manage interactive webcasts more effectively. Beyond effecting substantive changes to interactive webcasts for the course, the study may guide others who wish to pursue further studies of webcasts+chat as they occur in the course or elsewhere, or of other mixed-media environments, or who wish to adopt mixed-media environments for instruction. Other potential areas for research that emerged from this study include the affective states of participants in the webcasts+chat and the use of affective devices, such as emoticons and abbreviations, for showing affective states; the effect that format has on the efficacy of webcasts+chat used for computer-mediated instruction; and the processes students employ to manage actions and interactions in the webcasts and chats.Item Online experiment design with causal structures(2019-06-21) Sen, Rajat; Shakkottai, Sanjay; Caramanis, Constantine; Dimakis, Georgios-Alexandros; Johari, Ramesh; Sanghavi, SujayModern learning systems like recommendation engines, computational advertising systems, online parameter tuning services are inherently online; i.e. these systems need to continually collect data, take decisions to optimize a certain objective and then collect more data with the objective of improving their predictive abilities. This leads to the well-known exploration (searching the space of possible decisions) and exploitation (choosing the optimal decision according to the learned model) dilemma. A principled way to capture this trade-off is the study of multi-armed bandit problems. On the other hand, these online learning systems are made up of several interacting components. Therefore, it is beneficial to study the pattern of interaction among these components, in order to explore in a sample efficient manner, which in turn leads to better exploitation. In this thesis, we will see that it is sometimes beneficial to view these online learning systems under the lens of causality; thus formalizing the pattern of interaction among the various components of the system, through causal graphical models. In our first problem, we study the contextual bandit problem with L observed contexts and K arms, with a latent low dimensional causal structure. We show that leveraging this latent low dimensional structure can lead to superior regret guarantees that are practical even for smaller time horizons. This also leads to the first regret guarantees for low-rank matrix completion where the rank is greater than one. Our second problem deals with leveraging information leakage in an online fashion in the presence of causal structures. We identify that in presence of general causal structures there is information leakage between different interventions viewed as arms of a bandit i.e. collecting data under one intervention can inform us about the statistics under other interventions. We demonstrate how to leverage this information leakage through adaptive importance sampling and apply our algorithm in biological networks and for interpretability of deep networks. This directly leads us to our third problem, where we use the idea of information leakage (explored in our second problem) in the context of stochastic contextual bandits. We propose the contextual bandits with stochastic experts problem and provide the first problem dependent regret bound in contextual bandits, where the scaling of the regret bound can potentially be as low as logarithmic in the number of experts. We show that our algorithm outperforms several state of the art algorithms on progressive validation tasks on multi-class classification data-sets. In our fourth problem, we look at conditional independence testing, which is one of the fundamental tools in causal structure learning. We reduce this problem into binary classification, through a nearest neighbor based bootstrap procedure. This enables us to use powerful supervised learning tools like gradient boosted trees or deep neural networks, that have desirable properties in higher dimensions. Finally in our last two chapters, we explore the application of online learning techniques in the context of hyper-parameter tuning, which is of growing importance in general machine learning, as modern neural networks have several tunable parameters and even training one such parameter configuration can take several hours to days. In our sixth problem, we cast hyper-parameter tuning as optimizing a multi-fidelity black-box function (which is noise-less) and propose a multi-fidelity tree search algorithm for the same. In the seventh problem we extend our model and algorithm, so that they can function even in the presence of noise. We show that our tree-search based algorithms can outperform state of the art hyper-parameter tuning algorithms on several benchmark data-setsItem Online learning algorithms for wireless scheduling(2023-12) Song, Jianhan; De Veciana, Gustavo; Shakkottai , Sanjay; Hasenbein, John J; Mokhtari, Aryan; Caramanis, ConstantineOnline learning, and more specifically, multi-armed bandit algorithms, has recently garnered significant interest across diverse fields. Within an online learning framework, agents can leverage past interactions with their environment to optimize future decisions, making it an ideal mechanism for use in applications such as recommendation systems. Driven by these advantages, we believe that the online learning approach can be effectively employed to address resource allocation and scheduling challenges in wireless systems, with the potential to enhance the adaptability and robustness of system performance. In this dissertation, we explore the applications of multi-armed bandit algorithms in various wireless settings, showcasing their efficacy through both theoretical analysis and empirical demonstrations. We first studied the multi-user scheduling problem for the wireless downlink with instantaneous channel rate and queue information. We introduced the concept of "meta-scheduling", which formulates the task of selecting an optimal wireless scheduler as a bandit problem, and proposed a UCB-type bandit algorithm designed to adapt to the dynamics of a queueing system. Expanding on the meta-scheduling concept, we then studied a model of hierarchical scheduling in the context of network slicing, in which the base station learns the optimal option among infinitely-many arms. Our approach involves formulating the problem as a blackbox optimization and addressing it using an HOO-type bandit algorithm adaptive to random queueing cycles. Lastly, we transitioned into a multi-agent setting, where decisions of learning agents in close proximity are coupled with each other through interference. Within this context, we identified a low-complexity structure termed the "weakly-coupled system", and developed a decentralized bandit algorithm to facilitate the learning of optimal collective actions. Throughout each of these segments, we presented rigorous theoretical proofs demonstrating that the proposed algorithms exhibit the desired sub-linear regret compared to an idealized genie. Furthermore, we validated the efficacy of the algorithms through a series of experiments using simulation.Item Online learning and decision-making from implicit feedback(2017-05) Krishnasamy, Subhashini; Shakkottai, Sanjay; Vishwanath, Sriram; Baccelli, Francois; Zitkovic, Gordon; Srikant, RayadurgamThis thesis focuses on designing learning and control algorithms for emerging resource allocation platforms like recommender systems, 5G wireless networks, and online marketplaces. These systems have an environment which is only partially known. Thus, the controllers need to make resource allocation decisions based on implicit feedback obtained from the environment based on past actions. The goal is to sequentially select actions using incremental feedback so as to optimize performance while simultaneously learning about the environment. We study three problems which exemplify this setting. The first is an inference problem which requires identification of sponsored content in recommender systems. Specifically, we ask if it is possible to detect the existence of sponsored content disguised as genuine recommendations using implicit feedback from a subset of users of the recommender system. The second problem is the design of scheduling algorithms for switch networks when the user-server link statistics are unknown (for e.g., in wireless networks, online marketplaces). The scheduling algorithm has to tradeoff between scheduling the optimal links and obtaining sufficient feedback about all the links for accurate estimates. We observe the close connection of this problem to the stochastic multi-armed bandit problem and analyze bandit-style explore-exploit algorithms for learning the statistical parameters while simultaneously assigning servers to users. The third is the joint problem of base station activation and rate allocation in an energy efficient wireless network when the channel statistics are unknown. The controller observes instantaneous channel rates of activated BSs, and thereby sequentially obtains implicit feedback about the channel. Here again, there is a tradeoff between learning the channel versus optimizing the operation cost based on estimated parameters. For each of these systems, we propose algorithms with provable asymptotic guarantees. These learning algorithms highlight the use of implicit feedback in online decision making and control.Item Online learning for network resource allocation(2020-05) Basu, Soumya, Ph. D.; Nikolova, Evdokia; Shakkottai, Sanjay; Hasenbein, John J; De Veciana, Gustavo A; Sanghavi, SujayThe problem of efficiently allocating limited resources to competing demands manifests as the key problem in various real-world systems, such as road-traffic network, cellular network, content delivery network and so on. These problems have been studied for decades under the umbrella of network resource allocation, often positively impacting the design and operation of today’s network. However, with progress of technology, demand is increasing, the networks are becoming more complex and large, and new applications are emerging, creating a need for update to the existing solutions. In this thesis, we consider such problems in network resource allocation where new techniques are required for providing optimal solutions. Our central theme is to augment online learning techniques within existing solutions to facilitate improved and targeted operations. We study problems in, 1) road-traffic network, 2) content delivery network, 3) wireless networks, and 4) network of classifiers. In our first problem, we study the minimum toll-booth problem: placing minimum number of tollbooths to induce the original social optimal flow as the Nash equilibrium flow in the tolled network. We provide computational hardness results for the general congestion networks and provide polynomial time algorithm for series-parallel congestion networks. In our second problem, we design two adaptive time-to-live (TTL) caches which tunes its parameters automatically to achieve targeted performance. Specifically, we design dynamic TTL cache to provably achieve a feasible desired hit rate, while filtering TTL achieve the same desired hit rate with much less cache size utilization. All our theoretical results hold in the presence of non-stationary traffic. We empirically show the effectiveness of the proposed algorithms through simulations on 9 day web traffic trace from Akamai. In the third problem, we study cost optimization in a wireless network under stability constraints, where both base station (BS) switching and activation incur cost. Additionally, in our system BS activation requires learning the channel statistics as channels are observable only from active BSs. We design two algorithms which are explore-exploit in nature to solve this problem optimally. The key idea is to constraint BS switching and apply max- weight activation on switching instances. We design both static and queue-length dependent switching schemes, and show that the latter outperforms the former. In the fourth problem, we study an online unsupervised ensemble learning problem. In this problem, samples come to the system for classification in an online manner. The decision maker, across multiple rounds, routes the samples through a subset of classifiers from a fixed ensemble and collects labels from them. Finally, it aggregates the collected labels to provide a final label for the samples and complete its service with a guaranteed accuracy. We characterize the capacity region (in terms of accuracy level vs arrival rate) of this system, and provide a unsupervised learning aided algorithm which is throughput optimal in this setting. We validate our theory through experiments with two ensembles of deep neural networks used for image classification in grouped CIFAR-10 dataset. In our fifth problem, we study the expected reward maximization problem in the multi-armed bandit problem when each time an arm is played it gets blocked for a certain duration. In this problem, which we call Blocking Bandits, we show that the expected reward maximization problem with the knowledge of the mean rewards for each arm is computationally hard. The oracle greedy strategy that plays the available arm with the highest mean reward in each time slot provides an (1 − 1/e) fraction of the optimal reward. Furthermore, when the mean rewards are not known, playing the available arm with the highest upper confidence bound (ucb) index incurs a O(log(T)) regret w.r.t. the oracle greedy policy. We show that there are instances of the blocking bandits problem where Ω(log(T)) regret is unavoidable. Finally, we present both synthetic and real-data experiments to validate our theoretical claims.Item Online learning for scheduling in wireless networks(2022-05-06) Tariq, Isfar; Shakkottai, Sanjay; De Veciana, Gustavo A; Caramanis, Constantine; Baccelli, Francois; Hasenbein, John JOver the last few years, online learning has grown in importance as it allows us to build systems that can interact with the environment while continuously learning from past interactions to improve future decisions to maximize some objective. While online learning is used in several areas like recommendation systems, however, due to the complexity of wireless scheduling it is unclear how to utilize online learning. For instance, MU-MIMO scheduling involves the selection of a user subset and associated rate selection each time-slot for varying channel states (the vector of quantized channels matrices for each of the users) — a complex integer optimization problem that is different for each channel state. We propose that a low-dimensional structure is present in the wireless systems which can be exploited through online learning. For instance, channel-states "near" each other will likely have the same optimal solution. In our first problem, we present a framework through which we formulate the wireless scheduling problem as a multi-armed bandit problem. We then propose an online algorithm that can cluster the channel-states and learn the capacity region of these clusters. We show that our algorithms can significantly reduce the complexity of online learning for wireless settings and provide regret guarantees for our algorithm. In the second problem, we expand on our previous work and present (1) a framework that further exploits the low-dimensional structure present in the system by clustering users and (2) an online algorithm that utilizes the parameters learned by our previous algorithms to optimize the subset of users to be scheduled for given channel-state. We show that our algorithms can not only converge faster but also improve the overall throughput of the system. We also provide regret guarantees for the user clustering algorithm.Item Online teacher learning communities : how can Facebook support professional development?(2018-06-15) Mann, Deborah Mary; Marshall, Jill Ann; Borrego, Maura; Hobbs, Mary; Sampson, Victor; Keating, XiaofenDeveloping and supporting teacher identity has been largely overlooked in the professional development arena. Components of identity are typically associated with the more affective development of teaching capacities. This dissertation explored teacher identity as expressed on the Facebook page of a non-profit professional development organization, Ecorise. Teachers participated in face-to-face training with Ecorise and some were active participants of the private Facebook group. The Facebook page was used by the organization and teachers to share information and offer support. A framework of teacher identity was created from the literature. Four main categories included contextual identity, formative influences, professional identity and personal identity. Discourse analysis of the Facebook page interactions and analysis of survey responses highlighted use and impact of the Facebook page on and by teachers in the context of teacher identity. Findings in contextual identity focused in two main categories, discourse identity and affinity identity. The latter included indications of how the individuals aligned with the group identity. Liking and commenting on the Facebook page were gateways to moving to a more central rather than peripheral participation in the group. Facilitation of the group was not overt but occurred through group interaction, particularly validation. The nature of engagement differed between teachers new to the organization and those who were considered veterans. Some teachers reported preferring Canvas, an online course management tool focused on content delivery, as an online support, although many teachers were not aware of the Facebook page. Teachers using the Facebook page sought information on how to implement projects and topics. Facebook promotes reflection, an important formative influence. Through sharing stories teachers reflected on practice. These stories prompted validation providing encouragement to teachers. In exchanging “how-to’s” and telling stories teachers expressed and developed their pedagogical knowledge, the general focus being on instructional rather than content knowledge. Creativity was evident in the descriptions and photographs of student outcomes. Through all, the components of personal identity are intertwined illustrating the close connection between professional and personal identity and the concept of teaching as a calling. This emphasizes the importance of considering the affective domain in teacher training and investigating the use of tools, such as Facebook, that promote and support those qualities that build teacher identity.Item Parents learning online : informal education on parenting through online interactions examined from a community of practice perspective(2010-08) Matthews, Megan Renee; Schallert, Diane L.; Robinson, DanielThis study investigated the online interactions of parents using the constructs of Wenger’s (1998) community of practice theory. Parents were surveyed and blogs and comments selections were examined to determine whether a communities of practice perspective would be appropriate as a construct to examine parents’ online interactions, and whether parents could gain similar benefits to those found from face-to-face parent support groups. This study provides evidence to support the utility of parents’ online interactions and the relevance of a community of practice perspective as analyzed with the components of Wenger’s (1998) Communities of Practice Theory.Item The practitioner-driven system : an interactive qualitative analysis of the e-learning creation experience(2015-05) Derr, David Roy; Resta, Paul E.; McCoy, Danny; Hughes, Joan E.; Patterson, Jeffery; Riegle-Crumb, CatherineContemporary e-learning research often addresses a singular instructional topic, learning strategy, or authoring tool. Technological advancements and evolving delivery methods are changing the e-learning practitioner experience more rapidly than ever before, and the need for a holistic illustration of the modern-day practitioner experience has never been greater. This Interactive Qualitative Analysis (IQA) of the e-learning creation experience consists of Affinity Production Interviews, interviews, and an online survey of e-learning practitioners working with adult audiences. The result of the study is an e-learning creation experience system driven by participants’ stories. The system is comprised of twelve affinities including leadership, policy, the instructional systems design process, the client relationship, emotions, and more. Exercising the system reveals conditions that influence the ultimate outcome of the system: e-learning success.Item Provably efficient methods for large-scale learning(2023-07-19) Yang, Shuo, Ph. D.; Sanghavi, Sujay Rajendra, 1979-; Shakkottai, Sanjay; Caramanis, Constantine; Liu, QiangThe scale of machine learning problems grows rapidly in recent years and calls for efficient methods. In this dissertation, we propose simple and efficient methods for various large-scale learning problems. We start with a standard supervised learning problem of solving quadratic regression. In Chapter 2, we show that by utilizing the quadratic structure and a novel gradient estimation algorithm, we can solve sparse quadratic regression with sub-quadratic time complexity and near-optimal sample complexity. We then move to online learning problems. In Chapter 3, we identify a weak assumption and theoretically prove that the standard UCB algorithm efficiently learns from inconsistent human preferences with nearly optimal regret; in Chapter 4 we propose an approximate maximum inner product search data structure for adaptive queries and present two efficient algorithms that achieve sublinear time complexity for linear bandits, which is especially desirable for extremely large and slowly changing action sets. In Chapter 5, we study how to efficiently use privileged features with deep learning models. We present an efficient learning algorithm to exploit privileged features that are not available during testing time. We conduct comprehensive empirical evaluations and present rigorous analysis for linear models to build theoretical insights. It provides a general algorithmic paradigm that can be integrated with many other machine learning methods.