We study the problem of active learning with the added twist that the learner is assisted by a helpful teacher. We consider the following natural interaction protocol: At each round, the learner proposes a query asking for the label of an instance xq, the teacher provides the requested label {xq,yq} along with explanatory information to guide the learning process. In this paper, we view this information in the form of an additional contrastive example ({xc,yc}) where xc is picked from a set constrained by xq (e.g., dissimilar instances with the same label). Our focus is to design a teaching algorithm that can provide an informative sequence of contrastive examples to the learner to speed up the learning process. We show that this leads to a challenging sequence optimization problem where the algorithm's choices at a given round depend on the history of interactions. We investigate an efficient teaching algorithm that adaptively picks these contrastive examples. We derive strong performance guarantees for our algorithm based on two problem-dependent parameters and further show that for specific types of active learners (e.g., a generalized binary search learner), the proposed teaching algorithm exhibits strong approximation guarantees. Finally, we illustrate our bounds and demonstrate the effectiveness of our teaching framework via two numerical case studies.
more »
« less
Teaching Multiple Concepts to a Forgetful Learner
How can we help a forgetful learner learn multiple concepts within a limited time frame? While there have been extensive studies in designing optimal schedules for teaching a single concept given a learner's memory model, existing approaches for teaching multiple concepts are typically based on heuristic scheduling techniques without theoretical guarantees. In this paper, we look at the problem from the perspective of discrete optimization and introduce a novel algorithmic framework for teaching multiple concepts with strong performance guarantees. Our framework is both generic, allowing the design of teaching schedules for different memory models, and also interactive, allowing the teacher to adapt the schedule to the underlying forgetting mechanisms of the learner. Furthermore, for a well-known memory model, we are able to identify a regime of model parameters where our framework is guaranteed to achieve high performance. We perform extensive evaluations using simulations along with real user studies in two concrete applications: (i) an educational app for online vocabulary teaching; and (ii) an app for teaching novices how to recognize animal species from images. Our results demonstrate the effectiveness of our algorithm compared to popular heuristic approaches.
more »
« less
- Award ID(s):
- 1645832
- NSF-PAR ID:
- 10207092
- Editor(s):
- Wallach, H
- Date Published:
- Journal Name:
- Advances in neural information processing systems
- Volume:
- 32
- ISSN:
- 1049-5258
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The phenomenal growth in use of android devices in the recent years has also been accompanied by the rise of android malware. This reality warrants development of tools and techniques to analyze android apps in large scale for security vetting. Most of the state-of-the-art vetting tools are either based on static analysis or on dynamic analysis. Static analysis has limited success if the malware app utilizes sophisticated evading tricks. Dynamic analysis on the other hand may not find all the code execution paths, which let some malware apps remain undetected. Moreover, the existing static and dynamic analysis vetting techniques require extensive human interaction. To ad- dress the above issues, we design a deep learning based hybrid analysis technique, which combines the complementary strengths of each analysis paradigm to attain better accuracy. Moreover, automated feature engineering capability of the deep learning framework addresses the human interaction issue. In particular, using lightweight static and dynamic analysis procedure, we obtain multiple artifacts, and with these artifacts we train the deep learner to create independent models, and then combine them to build a hybrid classifier to obtain the final vetting decision (malicious apps vs. benign apps). The experiments show that our best deep learning model with hybrid analysis achieves an area under the precision-recall curve (AUC) of 0.9998. In this paper, we also present a comparative study of performance measures of the various variants of the deep learning framework. Additional experiments indicate that our vetting system is fairly robust against imbalanced data and is scalable.more » « less
-
We study the chance-constrained bin packing problem, with an application to hospital operating room planning. The bin packing problem allocates items of random sizes that follow a discrete distribution to a set of bins with limited capacity, while minimizing the total cost. The bin capacity constraints are satisfied with a given probability. We investigate a big-M and a 0-1 bilinear formulation of this problem. We analyze the bilinear structure of the formulation and use the lifting techniques to identify cover, clique, and projection inequalities to strengthen the formulation. We show that in certain cases these inequalities are facet-defining for a bilinear knapsack constraint that arises in the reformulation. An extensive computational study is conducted for the operating room planning problem that minimizes the number of open operating rooms. The computational tests are performed using problems generated based on real data from a hospital. A lower-bound improvement heuristic is combined with the cuts proposed in this paper in a branch-and-cut framework. The computations illustrate that the techniques developed in this paper can significantly improve the performance of the branch-and-cut method. Problems with up to 1,000 scenarios are solved to optimality in less than an hour. A safe approximation based on conditional value at risk (CVaR) is also solved. The computations show that the CVaR approximation typically leaves a gap of one operating room (e.g., six instead of five) to satisfy the chance constraint. Summary of Contribution: This paper investigates a branch-and-cut algorithm for a chance-constrained bin packing problem with multiple bins. The chance-constrained bin packing provides a modeling framework for applied operations research problems, such as health care, scheduling, and so on. This paper studies alternative computational approaches to solve this problem. Moreover, this paper uses real data from a hospital operating room planning setting as an application to test the algorithmic ideas. This work, therefore, is at the intersection of computing and operations research. Several interesting ideas are developed and studied. These include a strengthened big-M reformulation, analysis of a bilinear reformulation, and identifying certain facet-defining inequalities for this formulation. This paper also gives a lower-bound generation heuristic for a model that minimizes the number of bins. Computational experiments for an operating room planning model that uses data from a hospital demonstrate the computational improvement and importance of the proposed approaches. The techniques proposed in this paper and computational experiments further enhance the interface of computing and operations research.more » « less
-
Evidence has shown that facilitating student-centered learning (SCL) in STEM classrooms enhances student learning and satisfaction [1]–[3]. However, despite increased support from educational and government bodies to incorporate SCL practices [1], minimal changes have been made in undergraduate STEM curriculum [4]. Faculty often teach as they were taught, relying heavily on traditional lecture-based teaching to disseminate knowledge [4]. Though some faculty express the desire to improve their teaching strategies, they feel limited by a lack of time, training, and incentives [4], [5]. To maximize student learning while minimizing instructor effort to change content, courses can be designed to incorporate simpler, less time-consuming SCL strategies that still have a positive impact on student experience. In this paper, we present one example of utilizing a variety of simple SCL strategies throughout the design and implementation of a 4-week long module. This module focused on introductory tissue engineering concepts and was designed to help students learn foundational knowledge within the field as well as develop critical technical skills. Further, the module sought to develop important professional skills such as problem-solving, teamwork, and communication. During module design and implementation, evidence-based SCL teaching strategies were applied to ensure students developed important knowledge and skills within the short timeframe. Lectures featured discussion-based active learning exercises to encourage student engagement and peer collaboration [6]–[8]. The module was designed using a situated perspective, acknowledging that knowing is inseparable from doing [9], and therefore each week, the material taught in the two lecture sessions was directly applied to that week’s lab to reinforce students’ conceptual knowledge through hands-on activities and experimental outcomes. Additionally, the majority of assignments served as formative assessments to motivate student performance while providing instructors with feedback to identify misconceptions and make real-time module improvements [10]–[12]. Students anonymously responded to pre- and post-module surveys, which focused on topics such as student motivation for enrolling in the module, module expectations, and prior experience. Students were also surveyed for student satisfaction, learning gains, and graduate student teaching team (GSTT) performance. Data suggests a high level of student satisfaction, as most students’ expectations were met, and often exceeded. Students reported developing a deeper understanding of the field of tissue engineering and learning many of the targeted basic lab skills. In addition to hands-on skills, students gained confidence to participate in research and an appreciation for interacting with and learning from peers. Finally, responses with respect to GSTT performance indicated a perceived emphasis on a learner-centered and knowledge/community-centered approaches over assessment-centeredness [13]. Overall, student feedback indicated that SCL teaching strategies can enhance student learning outcomes and experience, even over the short timeframe of this module. Student recommendations for module improvement focused primarily on modifying the lecture content and laboratory component of the module, and not on changing the teaching strategies employed. The success of this module exemplifies how instructors can implement similar strategies to increase student engagement and encourage in-depth discussions without drastically increasing instructor effort to re-format course content. Introduction.more » « less
-
Calculation of many-body correlation functions is one of the critical kernels utilized in many scientific computing areas, especially in Lattice Quantum Chromodynamics (Lattice QCD). It is formalized as a sum of a large number of contraction terms each of which can be represented by a graph consisting of vertices describing quarks inside a hadron node and edges designating quark propagations at specific time intervals. Due to its computation- and memory-intensive nature, real-world physics systems (e.g., multi-meson or multi-baryon systems) explored by Lattice QCD prefer to leverage multi-GPUs. Different from general graph processing, many-body correlation function calculations show two specific features: a large number of computation-/data-intensive kernels and frequently repeated appearances of original and intermediate data. The former results in expensive memory operations such as tensor movements and evictions. The latter offers data reuse opportunities to mitigate the data-intensive nature of many-body correlation function calculations. However, existing graph-based multi-GPU schedulers cannot capture these data-centric features, thus resulting in a sub-optimal performance for many-body correlation function calculations. To address this issue, this paper presents a multi-GPU scheduling framework, MICCO, to accelerate contractions for correlation functions particularly by taking the data dimension (e.g., data reuse and data eviction) into account. This work first performs a comprehensive study on the interplay of data reuse and load balance, and designs two new concepts: local reuse pattern and reuse bound to study the opportunity of achieving the optimal trade-off between them. Based on this study, MICCO proposes a heuristic scheduling algorithm and a machine-learning-based regression model to generate the optimal setting of reuse bounds. Specifically, MICCO is integrated into a real-world Lattice QCD system, Redstar, for the first time running on multiple GPUs. The evaluation demonstrates MICCO outperforms other state-of-art works, achieving up to 2.25× speedup in synthesized datasets, and 1.49× speedup in real-world correlation functions.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
