Downloads & Free Reading Options - Results

Don't all AI methods need a bunch of data to work? How could AI help document and revitalize endangered languages with \"human-in-the-loop\" or \"active learning\" methods? Sarah Moeller from the University of Florida joins us to discuss those and other related questions. She also shares many of her personal experiences working with languages in low resource settings.

Don't all AI methods need a bunch of data to work? How could AI help document and revitalize endangered languages with \"human-in-the-loop\" or \"active learning\" methods? Sarah Moeller from the University of Florida joins us to discuss those and other related questions. She also shares many of her personal experiences working with languages in low resource settings.

pages

pages

pages

pages

pages

pages

Classroom response systems (CRS), also called student/audience response systems or clickers, have been used by business instructors, particularly in larger classes, to allow instructors to ask students questions in class and have their responses immediately tabulated and reported electronically. While clickers have typically been used to measure attendance, gauge comprehension and test students, we propose that classroom response systems can also be used to effectively create an active learning environment. Specifically we detail the use of classroom response systems to utilize active learning in large classes (i.e., more than 70 students) through describing five sample CRS exercises in Marketing courses: Sequential Elimination, Why Do You Think That?, Experiential Exercises, What Would You Do?, and Forced Choice. These exercises though, could be adapted for use in other business classes.

This study explored students and instructors' perceptions and experience of technology affordances in an technology-enhanced Active Learning Classroom (ALC) to promote students' collaborative problem solving. Multiple case studies were conducted. Five classes of 92 students and five professors participated in this study. The data sources were class observations, interviews, and pre- and post-surveys. The study showed that students' self-efficacy and confidence in completing problem-solving tasks increased over time. Additionally, it was found that some professors used the ALC to its potentials while others used it minimally. While both students and professors agreed about numerous benefits of ALCs for learning and instruction, how technology was used depended on the perceived purpose, needs and meaningfulness of the instructors. [For the full proceedings, see ED562107.]

This paper describes an Australian Learning and Teaching Council funded project for which Learning Design is encompassed in the broadest sense. ALIUS (Active Learning In University Science) takes the design of learning back to the learning experiences created for students. ALIUS is not about designing a particular activity, or subject, or course, but rather the development of a method, or process, by which we have re-designed the way in which learning occurs in large university classrooms world wide. (Contains 2 tables and 1 figure.)

This paper describes Project PLAI (Promoting Learning through Active Interaction), a 4-year research-to-practice project designed to develop and validate an early communication curriculum for infants who are deafblind and their caregivers. The curriculum provides a step-by-step approach to assist caregivers in recognizing their infants' early and subtle communication behaviors and in responding to these contingently, thus providing opportunity and encouragement for reciprocal interaction. This sequence offers an innovative format to assist caregivers in developing the ability to analyze and respond to early, subtle, and difficult-to-interpret communicative behaviors of their infants. This report discusses the conceptual framework of the Project PLAI model and its unique strategies, including its emphasis on the skills of both the caregivers and the infants, support of the natural efforts of caregivers, and the infusion of interactive strategies within the context of family routines, thus supporting caregiver-interactions in culturally responsive ways. The modules and objectives of the PLAI curriculum are outlined and include: (1) understanding infant cues; (2) identifying high and low preference objects, persons, and events; (3) establishing predictable routines; (4) establishing turn-taking; and (5) encouraging communicative initiations. Findings from Project PLAI are discussed and indicate strong support for the effectiveness of the PLAI curriculum. (CR)

Biology I students (n=166) in a high school in southern Mississippi completed the Learning Styles Inventory, a 45-item instrument designed to measure preference for channel of learning (visual, auditory, or kinesthetic), sociological environment (individually or in groups), and mode of expression (oral or written). A three-way analysis of variance was used to examine learning style preferences based on sex, race, and level of course (regular or honors). Students in the honors classes were significantly different from students in regular classes in preference for learning through visual language and for reporting their knowledge through oral expression. African-American females were significantly lower in preference for kinesthetic or active learning than were the other groups. There was a three-way interaction among sex, race, and level for written expression, with African American females and white males in honors classes showing higher preference for written expression than those in regular classes. There was no difference in preference for written expression between African American males and white females in honors and those in regular classes. A cluster analysis was used to describe three groups whose learning styles preferences can be addressed in the biology classroom: (1) the "scholars," who prefer learning through visual language, working individually, and showing what they have learned through written expression (21%); (2) the "active learners," who prefer learning through listening, engaging in active learning, and reporting what they have learned through oral expression (44%); and (3) "social butterflies," who show no preference other than working in groups (35%). (Contains 9 tables and 12 references.) (Author/SLD)

Contextual bandit algorithms -- a class of multi-armed bandit algorithms that exploit the contextual information -- have been shown to be effective in solving sequential decision making problems under uncertainty. A common assumption adopted in the literature is that the realized (ground truth) reward by taking the selected action is observed by the learner at no cost, which, however, is not realistic in many practical scenarios. When observing the ground truth reward is costly, a key challenge for the learner is how to judiciously acquire the ground truth by assessing the benefits and costs in order to balance learning efficiency and learning cost. From the information theoretic perspective, a perhaps even more interesting question is how much efficiency might be lost due to this cost. In this paper, we design a novel contextual bandit-based learning algorithm and endow it with the active learning capability. The key feature of our algorithm is that in addition to sending a query to an annotator for the ground truth, prior information about the ground truth learned by the learner is sent together, thereby reducing the query cost. We prove that by carefully choosing the algorithm parameters, the learning regret of the proposed algorithm achieves the same order as that of conventional contextual bandit algorithms in cost-free scenarios, implying that, surprisingly, cost due to acquiring the ground truth does not increase the learning regret in the long-run. Our analysis shows that prior information about the ground truth plays a critical role in improving the system performance in scenarios where active learning is necessary.

REVIEW

REVIEW

REVIEW

REVIEW

REVIEW

REVIEW

The use of theoretical classes that contextualize physiological concepts, followed by active methodologies has been common practice in our physiology classes.We employed active learning techniques in the medical course that involved physiology contents. We use brainstorming followed by conceptual maps, question-andanswer games, dramas, memory games, puzzle and portfolio creation.The goal of this assignment was to evaluate the active and passive methodologies in our educational context and to present a memory game – as active methodology - to explain the mechanisms of the electrical and chemical synapses.The game consisted of 2 boards and 40 foam pieces. The boards: one meant for electrical synapses and the other for chemical synapses, were divided into two columns to be fitted by the pieces: characteristics and the order of synaptic physiological events. The players were divided into two (n = 10) groups, one (n = 5) for each type of synapse. One player from each team chose two pieces simultaneously andthey were placed on the corresponding board.This procedure was repeated until one of the groups had filled the board with all the pieces related to the physiological characteristics of the synapses and to the sequences of synaptic events.At the end of the activity, it was concluded that learning the synaptic mechanisms through a playful process was more pleasant and fun compared to theoretical classes only and when strategically applied, active learning and passive learning are useful for teaching and for the learning process of medical school students.

We enhance the autonomy of the continuous active learning method shown by Cormack and Grossman (SIGIR 2014) to be effective for technology-assisted review, in which documents from a collection are retrieved and reviewed, using relevance feedback, until substantially all of the relevant documents have been reviewed. Autonomy is enhanced through the elimination of topic-specific and dataset-specific tuning parameters, so that the sole input required by the user is, at the outset, a short query, topic description, or single relevant document; and, throughout the review, ongoing relevance assessments of the retrieved documents. We show that our enhancements consistently yield superior results to Cormack and Grossman's version of continuous active learning, and other methods, not only on average, but on the vast majority of topics from four separate sets of tasks: the legal datasets examined by Cormack and Grossman, the Reuters RCV1-v2 subject categories, the TREC 6 AdHoc task, and the construction of the TREC 2002 filtering test collection.

We enhance the autonomy of the continuous active learning method shown by Cormack and Grossman (SIGIR 2014) to be effective for technology-assisted review, in which documents from a collection are retrieved and reviewed, using relevance feedback, until substantially all of the relevant documents have been reviewed. Autonomy is enhanced through the elimination of topic-specific and dataset-specific tuning parameters, so that the sole input required by the user is, at the outset, a short query, topic description, or single relevant document; and, throughout the review, ongoing relevance assessments of the retrieved documents. We show that our enhancements consistently yield superior results to Cormack and Grossman's version of continuous active learning, and other methods, not only on average, but on the vast majority of topics from four separate sets of tasks: the legal datasets examined by Cormack and Grossman, the Reuters RCV1-v2 subject categories, the TREC 6 AdHoc task, and the construction of the TREC 2002 filtering test collection.

We enhance the autonomy of the continuous active learning method shown by Cormack and Grossman (SIGIR 2014) to be effective for technology-assisted review, in which documents from a collection are retrieved and reviewed, using relevance feedback, until substantially all of the relevant documents have been reviewed. Autonomy is enhanced through the elimination of topic-specific and dataset-specific tuning parameters, so that the sole input required by the user is, at the outset, a short query, topic description, or single relevant document; and, throughout the review, ongoing relevance assessments of the retrieved documents. We show that our enhancements consistently yield superior results to Cormack and Grossman's version of continuous active learning, and other methods, not only on average, but on the vast majority of topics from four separate sets of tasks: the legal datasets examined by Cormack and Grossman, the Reuters RCV1-v2 subject categories, the TREC 6 AdHoc task, and the construction of the TREC 2002 filtering test collection.

Recent successes in learning-based image classification, however, heavily rely on the large number of annotated training samples, which may require considerable human efforts. In this paper, we propose a novel active learning framework, which is capable of building a competitive classifier with optimal feature representation via a limited amount of labeled training instances in an incremental learning manner. Our approach advances the existing active learning methods in two aspects. First, we incorporate deep convolutional neural networks into active learning. Through the properly designed framework, the feature representation and the classifier can be simultaneously updated with progressively annotated informative samples. Second, we present a cost-effective sample selection strategy to improve the classification performance with less manual annotations. Unlike traditional methods focusing on only the uncertain samples of low prediction confidence, we especially discover the large amount of high confidence samples from the unlabeled set for feature learning. Specifically, these high confidence samples are automatically selected and iteratively assigned pseudo-labels. We thus call our framework "Cost-Effective Active Learning" (CEAL) standing for the two advantages.Extensive experiments demonstrate that the proposed CEAL framework can achieve promising results on two challenging image classification datasets, i.e., face recognition on CACD database [1] and object categorization on Caltech-256 [2].

Active learning is an important machine learning problem in reducing the human labeling effort. Current active learning strategies are designed from human knowledge, and are applied on each dataset in an immutable manner. In other words, experience about the usefulness of strategies cannot be updated and transferred to improve active learning on other datasets. This paper initiates a pioneering study on whether active learning experience can be transferred. We first propose a novel active learning model that linearly aggregates existing strategies. The linear weights can then be used to represent the active learning experience. We equip the model with the popular linear upper- confidence-bound (LinUCB) algorithm for contextual bandit to update the weights. Finally, we extend our model to transfer the experience across datasets with the technique of biased regularization. Empirical studies demonstrate that the learned experience not only is competitive with existing strategies on most single datasets, but also can be transferred across datasets to improve the performance on future learning tasks.

We develop a new active learning algorithm for the streaming setting satisfying three important properties: 1) It provably works for any classifier representation and classification problem including those with severe noise. 2) It is efficiently implementable with an ERM oracle. 3) It is more aggressive than all previous approaches satisfying 1 and 2. To do this we create an algorithm based on a newly defined optimization problem and analyze it. We also conduct the first experimental analysis of all efficient agnostic active learning algorithms, evaluating their strengths and weaknesses in different settings.

We present a new active learning algorithm based on nonparametric estimators of the regression function. Our investigation provides probabilistic bounds for the rates of convergence of the generalization error achievable by proposed method over a broad class of underlying distributions. We also prove minimax lower bounds which show that the obtained rates are almost tight.

We present a new active learning algorithm based on nonparametric estimators of the regression function. Our investigation provides probabilistic bounds for the rates of convergence of the generalization error achievable by proposed method over a broad class of underlying distributions. We also prove minimax lower bounds which show that the obtained rates are almost tight.

We present a new active learning algorithm based on nonparametric estimators of the regression function. Our investigation provides probabilistic bounds for the rates of convergence of the generalization error achievable by proposed method over a broad class of underlying distributions. We also prove minimax lower bounds which show that the obtained rates are almost tight.

We present a new active learning algorithm based on nonparametric estimators of the regression function. Our investigation provides probabilistic bounds for the rates of convergence of the generalization error achievable by proposed method over a broad class of underlying distributions. We also prove minimax lower bounds which show that the obtained rates are almost tight.

We guess humans start acquiring grasping skills as early as at the infant stage by virtue of two key processes. First, infants attempt to learn grasps for known objects by imitating humans. Secondly, knowledge acquired during this process is reused in learning to grasp novel objects. We argue that these processes of active and transfer learning boil down to a random search of grasps on an object, suitably biased by prior experience. In this paper we introduce active learning of grasps for known objects as well as transfer learning of grasps for novel objects grounded on kernel adaptive, mode-hopping Markov Chain Monte Carlo. Our experiments show promising applicability of our proposed learning methods.

The purpose of this study is to examine the effect that using active learning techniques during museum and gallery visits has on teacher candidates' academic success rates in and attitudes toward their Visual Arts Course. In this study, the importance and requirement of education to take place in museums and art galleries is emphasized. The research was conducted during the fall term of 4th year students' 2013-2014 academic in the Visual Arts Education course of Ataturk University's Kazim Karabekir School of Education's School Teaching Art Education Department. The present study had a semi-experimental design, included an experimental group (n = 30) and a control group (n = 30), and used pretest and posttest measures. Semi-experimental pattern's pretest-posttest test control group model were preferred in this study, realized by using a combined method research. In this study, realized by using a combined method research, semi-structured interview forms were used to obtain the qualitative data. Sixty 4th-year teacher candidates studying in the School Teaching Department participated in the research. As the result of the research, it was observed that museums and art galleries have positive effects on teacher candidates' attitudes toward active learning.

Turning lectures into interactive, student-led question and answer sessions is known to increase learning, but enabling interaction in a large class seems an insurmountable task. This can discourage adoption of this new approach -- who has time to individualize responses, address questions from over 200 students and encourage active participation in class? An approach adopted by a teaching team in large first-year classes at a research-intensive university appears to provide a means to do so. We describe the implementation of active learning strategies in a large first-year undergraduate physics unit of study, replacing traditional, content-heavy lectures with an integrated approach to question-driven learning. A key feature of our approach is that it facilitates intensive in-class discussions by requiring students to engage in preparatory reading and answer short written quizzes before every class. The lecturer uses software to rapidly analyze the student responses and identify the main issues faced by the students before the start of each class. We report the success of the integration of student preparation with this analysis and feedback framework, and the impact on the in-class discussions. We also address some of the difficulties commonly experienced by staff preparing for active learning classes.

Active learning is a student's active impact on learning and a student's involvement in the learning process which allows students to focus on creating knowledge with an emphasis on skills such as analytical thinking, problem-solving and meta-cognitive activities that develop students' thinking. The main purpose of this study is to determine whether there is a significant difference in teaching science class between the average attitude scores of active learning group and those of traditional learning group. "Pretest-posttest control group design" has been used within the study. The subjects of the groups have been 7th grade students in spring term of 2015-2016 school year in Melahat Unugur Secondary School in the city centre of Eskisehir. Two similar classes have been formed by analysing average scores from pretest and pre-attitude and 7-C and 7-D classes have been randomly selected as experimental and control groups, respectively. While the education program and course materials have been prepared and utilised with regard to active learning method in the experimental group, traditional method has been used in the control group. There has been a significant difference between experimental and control groups in favour of experimental group at the end of the study.

Metacognitive awareness constitutes a part of thinking skills to continuously retain and develop. One of efforts to increase metacognitive awareness is conducting training through lecture integrated in active learning. Self and peer-assessments encourage students to be more responsible for their performance improvement and learning. The population comprised students attending Natural Science course in Faculty of Teacher Training and Education, University of Muhammadiyah Malang; while the samples were all members of the population taken by means of saturation sampling technique. The technique of data collection was by means of Schraw and Dennison MAI (Metacognition Awareness Instrument), written test, and document analysis. The result path analysis showed: 1) there are influences of self and peer assessment in active learning towards awareness of Metacognition and cognitive ability; 2) cognitive knowledge, regulation of cognitive, and combined both contributed to its influence on Metacognition awareness; 3) cognitive knowledge, regulation of cognitive, and combined both contributed to its influence on cognitive variables; and 4) metacognitive awareness variables have a direct contribution to cognitive variables that are the influence of factors outside the metacognitive awareness variable.

active learning

Economic development has encouraged innovation in a range of industries; however, the education industry has been following traditional modes of teaching until recently across the world. This has not only affected the capabilities of the youth but has also led to the problem of unemployment and diminished success rates at the workplace. One of the factors contributing to this situation is the lack of student engagement. In this research, we examined the factors affecting student engagement among higher education students and determining the impact of pedagogical approaches with active leaning on influencing the engagement level. Analysis of 50 higher education level 3-7 students has shown that the major hinderance in student engagement is assessing their current knowledge levels to teach lessons effectively. Hence, using practical knowledge impartation mechanisms such as gamification, collaborative and peer learning would improve the cognitive, learning, and emotional behaviour of students; thus, boosting the engagement level.

xii, 240 p. : 28 cm

This paper addresses the problem of active learning of a multi-output Gaussian process (MOGP) model representing multiple types of coexisting correlated environmental phenomena. In contrast to existing works, our active learning problem involves selecting not just the most informative sampling locations to be observed but also the types of measurements at each selected location for minimizing the predictive uncertainty (i.e., posterior joint entropy) of a target phenomenon of interest given a sampling budget. Unfortunately, such an entropy criterion scales poorly in the numbers of candidate sampling locations and selected observations when optimized. To resolve this issue, we first exploit a structure common to sparse MOGP models for deriving a novel active learning criterion. Then, we exploit a relaxed form of submodularity property of our new criterion for devising a polynomial-time approximation algorithm that guarantees a constant-factor approximation of that achieved by the optimal set of selected observations. Empirical evaluation on real-world datasets shows that our proposed approach outperforms existing algorithms for active learning of MOGP and single-output GP models.

Active learning strategies respond to the costly labelling task in a supervised classification by selecting the most useful unlabelled examples in training a predictive model. Many conventional active learning algorithms focus on refining the decision boundary, rather than exploring new regions that can be more informative. In this setting, we propose a sequential algorithm named EG-Active that can improve any Active learning algorithm by an optimal random exploration. Experimental results show a statistically significant and appreciable improvement in the performance of our new approach over the existing active feedback methods.

Applications that learn from opinionated documents, like tweets or product reviews, face two challenges. First, the opinionated documents constitute an evolving stream, where both the author's attitude and the vocabulary itself may change. Second, labels of documents are scarce and labels of words are unreliable, because the sentiment of a word depends on the (unknown) context in the author's mind. Most of the research on mining over opinionated streams focuses on the first aspect of the problem, whereas for the second a continuous supply of labels from the stream is assumed. Such an assumption though is utopian as the stream is infinite and the labeling cost is prohibitive. To this end, we investigate the potential of active stream learning algorithms that ask for labels on demand. Our proposed ACOSTREAM 1 approach works with limited labels: it uses an initial seed of labeled documents, occasionally requests additional labels for documents from the human expert and incrementally adapts to the underlying stream while exploiting the available labeled documents. In its core, ACOSTREAM consists of a MNB classifier coupled with "sampling" strategies for requesting class labels for new unlabeled documents. In the experiments, we evaluate the classifier performance over time by varying: (a) the class distribution of the opinionated stream, while assuming that the set of the words in the vocabulary is fixed but their polarities may change with the class distribution; and (b) the number of unknown words arriving at each moment, while the class polarity may also change. Our results show that active learning on a stream of opinionated documents, delivers good performance while requiring a small selection of labels

We investigate active learning with access to two distinct oracles: Label (which is standard) and Search (which is not). The Search oracle models the situation where a human searches a database to seed or counterexample an existing solution. Search is stronger than Label while being natural to implement in many situations. We show that an algorithm using both oracles can provide exponentially large problem-dependent improvements over Label alone.

In our today's information society more and more data emerges, e.g.~in social networks, technical applications, or business applications. Companies try to commercialize these data using data mining or machine learning methods. For this purpose, the data are categorized or classified, but often at high (monetary or temporal) costs. An effective approach to reduce these costs is to apply any kind of active learning (AL) methods, as AL controls the training process of a classifier by specific querying individual data points (samples), which are then labeled (e.g., provided with class memberships) by a domain expert. However, an analysis of current AL research shows that AL still has some shortcomings. In particular, the structure information given by the spatial pattern of the (un)labeled data in the input space of a classification model (e.g.,~cluster information), is used in an insufficient way. In addition, many existing AL techniques pay too little attention to their practical applicability. To meet these challenges, this article presents several techniques that together build a new approach for combining AL and semi-supervised learning (SSL) for support vector machines (SVM) in classification tasks. Structure information is captured by means of probabilistic models that are iteratively improved at runtime when label information becomes available. The probabilistic models are considered in a selection strategy based on distance, density, diversity, and distribution (4DS strategy) information for AL and in a kernel function (Responsibility Weighted Mahalanobis kernel) for SVM. The approach fuses generative and discriminative modeling techniques. With 20 benchmark data sets and with the MNIST data set it is shown that our new solution yields significantly better results than state-of-the-art methods.

In our today's information society more and more data emerges, e.g.~in social networks, technical applications, or business applications. Companies try to commercialize these data using data mining or machine learning methods. For this purpose, the data are categorized or classified, but often at high (monetary or temporal) costs. An effective approach to reduce these costs is to apply any kind of active learning (AL) methods, as AL controls the training process of a classifier by specific querying individual data points (samples), which are then labeled (e.g., provided with class memberships) by a domain expert. However, an analysis of current AL research shows that AL still has some shortcomings. In particular, the structure information given by the spatial pattern of the (un)labeled data in the input space of a classification model (e.g.,~cluster information), is used in an insufficient way. In addition, many existing AL techniques pay too little attention to their practical applicability. To meet these challenges, this article presents several techniques that together build a new approach for combining AL and semi-supervised learning (SSL) for support vector machines (SVM) in classification tasks. Structure information is captured by means of probabilistic models that are iteratively improved at runtime when label information becomes available. The probabilistic models are considered in a selection strategy based on distance, density, diversity, and distribution (4DS strategy) information for AL and in a kernel function (Responsibility Weighted Mahalanobis kernel) for SVM. The approach fuses generative and discriminative modeling techniques. With 20 benchmark data sets and with the MNIST data set it is shown that our new solution yields significantly better results than state-of-the-art methods.

In our today's information society more and more data emerges, e.g.~in social networks, technical applications, or business applications. Companies try to commercialize these data using data mining or machine learning methods. For this purpose, the data are categorized or classified, but often at high (monetary or temporal) costs. An effective approach to reduce these costs is to apply any kind of active learning (AL) methods, as AL controls the training process of a classifier by specific querying individual data points (samples), which are then labeled (e.g., provided with class memberships) by a domain expert. However, an analysis of current AL research shows that AL still has some shortcomings. In particular, the structure information given by the spatial pattern of the (un)labeled data in the input space of a classification model (e.g.,~cluster information), is used in an insufficient way. In addition, many existing AL techniques pay too little attention to their practical applicability. To meet these challenges, this article presents several techniques that together build a new approach for combining AL and semi-supervised learning (SSL) for support vector machines (SVM) in classification tasks. Structure information is captured by means of probabilistic models that are iteratively improved at runtime when label information becomes available. The probabilistic models are considered in a selection strategy based on distance, density, diversity, and distribution (4DS strategy) information for AL and in a kernel function (Responsibility Weighted Mahalanobis kernel) for SVM. The approach fuses generative and discriminative modeling techniques. With 20 benchmark data sets and with the MNIST data set it is shown that our new solution yields significantly better results than state-of-the-art methods.

In our today's information society more and more data emerges, e.g.~in social networks, technical applications, or business applications. Companies try to commercialize these data using data mining or machine learning methods. For this purpose, the data are categorized or classified, but often at high (monetary or temporal) costs. An effective approach to reduce these costs is to apply any kind of active learning (AL) methods, as AL controls the training process of a classifier by specific querying individual data points (samples), which are then labeled (e.g., provided with class memberships) by a domain expert. However, an analysis of current AL research shows that AL still has some shortcomings. In particular, the structure information given by the spatial pattern of the (un)labeled data in the input space of a classification model (e.g.,~cluster information), is used in an insufficient way. In addition, many existing AL techniques pay too little attention to their practical applicability. To meet these challenges, this article presents several techniques that together build a new approach for combining AL and semi-supervised learning (SSL) for support vector machines (SVM) in classification tasks. Structure information is captured by means of probabilistic models that are iteratively improved at runtime when label information becomes available. The probabilistic models are considered in a selection strategy based on distance, density, diversity, and distribution (4DS strategy) information for AL and in a kernel function (Responsibility Weighted Mahalanobis kernel) for SVM. The approach fuses generative and discriminative modeling techniques. With 20 benchmark data sets and with the MNIST data set it is shown that our new solution yields significantly better results than state-of-the-art methods.

We propose a new active learning (AL) method for text classification with convolutional neural networks (CNNs). In AL, one selects the instances to be manually labeled with the aim of maximizing model performance with minimal effort. Neural models capitalize on word embeddings as representations (features), tuning these to the task at hand. We argue that AL strategies for multi-layered neural models should focus on selecting instances that most affect the embedding space (i.e., induce discriminative word representations). This is in contrast to traditional AL approaches (e.g., entropy-based uncertainty sampling), which specify higher level objectives. We propose a simple approach for sentence classification that selects instances containing words whose embeddings are likely to be updated with the greatest magnitude, thereby rapidly learning discriminative, task-specific embeddings. We extend this approach to document classification by jointly considering: (1) the expected changes to the constituent word representations; and (2) the model's current overall uncertainty regarding the instance. The relative emphasis placed on these criteria is governed by a stochastic process that favors selecting instances likely to improve representations at the outset of learning, and then shifts toward general uncertainty sampling as AL progresses. Empirical results show that our method outperforms baseline AL approaches on both sentence and document classification tasks. We also show that, as expected, the method quickly learns discriminative word embeddings. To the best of our knowledge, this is the first work on AL addressing neural models for text classification.

We propose a new active learning (AL) method for text classification with convolutional neural networks (CNNs). In AL, one selects the instances to be manually labeled with the aim of maximizing model performance with minimal effort. Neural models capitalize on word embeddings as representations (features), tuning these to the task at hand. We argue that AL strategies for multi-layered neural models should focus on selecting instances that most affect the embedding space (i.e., induce discriminative word representations). This is in contrast to traditional AL approaches (e.g., entropy-based uncertainty sampling), which specify higher level objectives. We propose a simple approach for sentence classification that selects instances containing words whose embeddings are likely to be updated with the greatest magnitude, thereby rapidly learning discriminative, task-specific embeddings. We extend this approach to document classification by jointly considering: (1) the expected changes to the constituent word representations; and (2) the model's current overall uncertainty regarding the instance. The relative emphasis placed on these criteria is governed by a stochastic process that favors selecting instances likely to improve representations at the outset of learning, and then shifts toward general uncertainty sampling as AL progresses. Empirical results show that our method outperforms baseline AL approaches on both sentence and document classification tasks. We also show that, as expected, the method quickly learns discriminative word embeddings. To the best of our knowledge, this is the first work on AL addressing neural models for text classification.

In many real world problems, control decisions have to be made with limited information. The controller may have no a priori (or even posteriori) data on the nonlinear system, except from a limited number of points that are obtained over time. This is either due to high cost of observation or the highly non-stationary nature of the system. The resulting conflict between information collection (identification, exploration) and control (optimization, exploitation) necessitates an active learning approach for iteratively selecting the control actions which concurrently provide the data points for system identification. This paper presents a dual control approach where the information acquired at each control step is quantified using the entropy measure from information theory and serves as the training input to a state-of-the-art Gaussian process regression (Bayesian learning) method. The explicit quantification of the information obtained from each data point allows for iterative optimization of both identification and control objectives. The approach developed is illustrated with two examples: control of logistic map as a chaotic system and position control of a cart with inverted pendulum.

LibriVox volunteers bring you fifteen different readings of <i>Mother Night</i>, by James Weldon Johnson. This weekly poetry project (for the week of Feb 26, 2006) was selected to celebrate Black History Month.<br>(Summary by Annie Coleman)

In celebration of <a href="http://en.wikipedia.org/wiki/Juneteenth">Juneteenth</a>, LibriVox volunteers bring you five different versions of <em>O Southland!</em>, by James Weldon Johnson. This was the weekly poetry project for 18 June 2006.<br />(Summary by Annie Coleman)

3Song My Paddle Sings

By E. Pauline Johnson

LibriVox’s weekly poetry project for the week of February 5, 2006 offers fourteen versions of "The Song My Paddle Sings" from the collection <i>Flint and Feather</i> by E. Pauline Johnson.<br /><br /> E. Pauline Johnson, also known as Tekahionwake, was born to the Mohawk Chief G.H.M. Johnson (Onwanonsyshon), and his wife, Emily S. Howells, a lady of pure English parentage. Pauline, born and raised in Canada, was a great reader and began writing poetry as a child. She died in 1913 after having poetry published in periodicals in several countries and collections of her work published in book form. (Summary adapted from the "Biographical Sketch" included in the Gutenberg text.)

4Plan and Preface to a Dictionary of the English Language

By Samuel Johnson

The published dictionary was a huge book: with pages nearly 1½ feet tall and 20 inches wide, it contained 42,773 words; it also sold for the huge price of £4/10s. ($400?). It would be years before "Johnson's Dictionary", as it came to be known, would ever turn a profit; authors' royalties being unknown at that time, Johnson, once his contract to deliver the book was fulfilled, received no further monies connected to the book.<br><br> Johnson, once again a freelance writer, albeit now a famous one, faced a grim hand-to-mouth existence; however, in July 1762 the twenty-four year old King George III granted Johnson an annual pension of £300. While not making Johnson rich, it allowed him a modest yet comfortable independence for the remaining thirty years of his life. (Summary from Wikipedia).

A book of simple receipts for little folk with important cooking rules in rhyme together with handy lists of the materials and utensils needed for the preparation of each dish. (Summary from the text)

6Rasselas, Prince of Abyssinia

By Samuel Johnson

In this enchanting fable (subtitled The Choice of Life), Rasselas and his retinue burrow their way out of the totalitarian paradise of the Happy Valley in search of that triad of eighteenth-century aspiration - life, liberty and happiness.<br><br/>According to that quirky authority, James Boswell, Johnson penned his only work of prose fiction in a handful of days to cover the cost of his mother's funeral. The stylistic elegance of the book and its wide-ranging philosophical concerns give no hint of haste or superficiality. <br><br/> Among other still burning issues Johnson's characters pursue questions of education, colonialism, the nature of the soul and even climate alteration. <br> Johnson's profoundest concern, however, is with the alternating attractions of solitude and social participation, seen not only as the ultimate life-choice but as the arena in which are played out the deepest fears of the individual: "Of the uncertainties of our present state, the most dreadful and alarming is the uncertain continuance of Reason.” (Summary by Martin Geeson)

A General History of the Robberies and Murders of the most notorious Pyrates is a 1724 book containing biographies of contemporary pirates. It's author uses the name Captain Charles Johnson, generally considered a pseudonym.<br /><br /> The real identity of the author was thought by some scholars to be Daniel Defoe, although this has since been disputed. The publisher Nathaniel Mist or somebody working for him are other suggested authors.<br /><br /> In the first volume, "Johnson" sticks fairly close to the available sources, though he embellishes the stories somewhat. Nevertheless, the book was influential in shaping popular notions of piracy and provided the standard account of the lives of many individuals still famous in the 21st century.(Summary by Wikipedia)

A year in the life of a New England farm boy at the end of the 19th century (Introduction by LC)

9Train Dogs

By E. Pauline Johnson

LibriVox volunteers bring you 18 recordings of The Train Dogs by E. Pauline Johnson. This was the Weekly Poetry project for February 6th, 2011.<br><br>Emily Pauline Johnson (Mohawk: Tekahionwake –pronounced: dageh-eeon-wageh, literally: 'double-life')(10 March 1861 – 7 March 1913), commonly known as E. Pauline Johnson or just Pauline Johnson, was a Canadian writer and performer popular in the late 19th century. Johnson was notable for her poems and performances that celebrated her First Nations heritage; she also had half English ancestry. One such poem is the frequently anthologized "The Song My Paddle Sings". Her poetry was published in Canada, the United States and Great Britain. Johnson was one of a generation of widely read writers who began to define a Canadian literature.

10Sunset in the Tropics

By James Weldon Johnson

LibriVox volunteers bring you 14 recordings of "Sunset in the Tropics." This is the Weekly Poetry for the week of August 10, 2014.<br><br>The author of this poem, James Weldon Johnson, served as U. S. Consul to Venezuela and Nicaragua, was an early leader in the NAACP and contributed to the Harlem Renaissance of the 1920s. He had a broad appreciation for black artists, musicians and writers, and worked to heighten awareness of their creativity. (from Wikipedia)

A shot penetrates the still night air and one of Galton's leading citizens is a victim of a desperate killer's gun. Murder becomes entangled with politics and it takes a parish priest to discover "whodunit" in this thriller by a husband and wife novelist team. (Summary taken from original jacket blurb.)

County Prosecutor Wally Brighton was found shot to death one evening, a broken rosary in one hand and a .32 automatic inches away from the other. Was it murder or suicide? It takes two sharp-eyed reporters, who combine romance with amateur sleuthing, to find the surprising solution to this baffling and thrilling murder mystery. (Summary from an original jacket)

13Cry From An Indian Wife

By E. Pauline Johnson

LibriVox volunteers bring you 13 recordings of A Cry From an Indian Wife by E. Pauline Johnson,. This was the Fortnightly Poetry project for January 29, 2012.<br><br>In 1892 the opportunity of a lifetime came to this young versifier, when Frank Yeigh, the president of the Young Liberals' Club, of Toronto, conceived the idea of having an evening of Canadian literature, at which all available Canadian authors should be guests and read from their own works.<br><br>Among the authors present on this occasion was Pauline Johnson, who contributed to the programme one of her compositions, entitled "A Cry from an Indian Wife"; and when she recited without text this much-discussed poem, which shows the Indian's side of the North-West Rebellion, she was greeted with tremendous applause from an audience which represented the best of Toronto's art, literature and culture. She was the only one on the programme who received an encore, and to this she replied with one of her favourite canoeing poems.<br><br>The following morning the entire press of Toronto asked why this young writer was not on the platform as a professional reader; while two of the dailies even contained editorials on the subject, inquiring why she had never published a volume of her poems, and insisted so strongly that the public should hear more of her, that Mr. Frank Yeigh arranged for her to give an entire evening in Association Hall within two weeks from the date of her first appearance. It was for this first recital that she wrote the poem by which she is best known, "The Song my Paddle Sings."<br>( Summary from the Biographical Sketch included in Flint And Feather, collected verse by E. Pauline Johnson )

14Indian Corn Planter

By E. Pauline Johnson

LibriVox volunteers bring you 13 recordings of The Indian Corn Planter by E. Pauline Johnson. This was the Weekly Poetry project for April 29, 2012.<br><br><i>"But in the writings of one poet alone I came upon a new note—the note of the Red Man's Canada. This was the poet that most interested me—Pauline Johnson. I quoted her lovely canoe song "In the Shadows," which will be found in this volume. I at once sat down and wrote a long article, which could have been ten times as long, upon a subject so suggestive as that of Canadian poetry."</i> (From the Introduction to Flint and Feather, Collected Verse BY E. Pauline Johnson; written by Theodore Watts-Dunton, The Pines, Putney Hill. 20th August, 1913.

15Fire - Flowers

By E. Pauline Johnson

LibriVox volunteers bring you 17 recordings of Fire - Flowers by E. Pauline Johnson. This was the Fortnightly Poetry project for August 18, 2013.<br><br>Fire-Flowers is taken from the book, Flint and Feather: Collected Verse by E. Pauline Johnson. (Summary by David Lawrence)

16Word of an Engineer

By James Weldon Johnson

James Weldon Johnson was an American author, educator, lawyer, diplomat, songwriter, and civil rights activist. Johnson is best remembered for his leadership within the National Association for the Advancement of Colored People (NAACP), where he started working in 1917, being chosen as the first black executive secretary of the organization, effectively the operating officer. He was first known for his writing, which includes poems, novels, and anthologies collecting both poems and spirituals of black culture. (Summary from Wikipedia)

17Morning, Noon and Night

By James Weldon Johnson

LibriVox volunteers bring you 14 readings of Morning, Noon and Night by James Weldon Johnson. This was the weekly poetry project for the week of September 14th, 2014.<br /><br />James Weldon Johnson served as U. S. Consul to Venezuela and Nicaragua, was an early leader in the NAACP and contributed to the Harlem Renaissance of the 1920s. He had a broad appreciation for black artists, musicians and writers, and worked to heighten awareness of their creativity. (from Wikipedia)

This is a volume of selected essays by "the great master of reason" Samuel Johnson. The most famous exerpts from The Rambler, The Adventurer and The Idler are included, covering a vast range of topics. - Summary by Carolin

19Deep In The Quiet Wood

By James Weldon Johnson

James Weldon Johnson was an American author, educator, lawyer, diplomat, songwriter, and civil rights activist. Johnson is best remembered for his leadership of the National Association for the Advancement of Colored People (NAACP), where he started working in 1917. In 1920 he was the first black to be chosen as executive secretary of the organization, effectively the operating officer. He served in that position from 1920 to 1930. Johnson established his reputation as a writer, and was known during the Harlem Renaissance for his poems, novels, and anthologies collecting both poems and spirituals of black culture.

20Sonnet (From the Spanish of Plácido)

By James Weldon Johnson

James Weldon Johnson was an American author, educator, lawyer, diplomat, songwriter, and civil rights activist. Johnson is best remembered for his leadership of the National Association for the Advancement of Colored People (NAACP), where he started working in 1917. In 1920 he was the first black to be chosen as executive secretary of the organization, effectively the operating officer. - Summary by Wikipedia

21Autobiography of an Ex-Colored Man (version 2)

By James Weldon Johnson

The story of a biracial man living in the deep south after the reconstruction era. He is young and talented. Yet, in order for him to avoid stigma, he has to pass as white. But would his sense of belonging and solidarity to the black community of the time catch up with him at the end? "It is very likely that the Negroes of the United States have a fairly correct idea of what the white people of the country think of them, for that opinion has for a long time been and is still being constantly stated; but they are themselves more or less a sphinx to the whites. It is curiously interesting and even vitally important to know what are the thoughts of ten millions of them concerning the people among whom they live. In these pages it is as though a veil had been drawn aside: the reader is given a view of the inner life of the Negro in America, is initiated into the "freemasonry," as it were, of the race. These pages also reveal the unsuspected fact that prejudice against the Negro is exerting a pressure which, in New York and other large cities where the opportunity is open, is actually and constantly forcing an unascertainable number of fair-complexioned colored people over into the white race. In this book the reader is given a glimpse behind the scenes of this race-drama which is being here enacted,—he is taken upon an elevation where he can catch a bird's-eye view of the conflict which is being waged." (from the preface by the publisher and Stav Nisser.)

22Autobiography of an Ex-Colored Man

By James Weldon Johnson

Johnson's only novel, <i>The Autobiography of an Ex-Colored Man</i>, was originally published anonymously in 1912. It is a fictional novel written as a memoir of an unnamed biracial narrator who grew up in the South during the Reconstruction and post-Reconstruction eras. It is a story in which the narrator relates how as a young boy he initially assumed that he was white, and how his notions of racial identity were suddenly turned upside down one day—how from that moment on he was inclined to view himself and the world about him from the perspective of blackness. The novel received very little notoriety until Johnson republished it in 1927, this time taking full credit as its author. - Summary by James K. White

23Camper

By E. Pauline Johnson

It is eminently fitting that this daughter of Nature should have been laid to rest in no urban cemetery. According to her own request she was buried in Stanley Park, Vancouver's beautiful heritage of the forest primeval. A simple stone surrounded by rustic palings marks her grave and on this stone is carved the one word "Pauline." There she lies among ferns and wild flowers a short distance from Siwash Rock, the story of which she has recorded in the legends of her race. In time to come a pathway to her grave will be worn by lovers of Canadian poetry who will regard it as one of the most romantic of our literary shrines. (from the Biographical Sketch in Flint and Feather: Collected Verse by E. Pauline Johnson)

24Homing Bee

By E. Pauline Johnson

It is eminently fitting that this daughter of Nature should have been laid to rest in no urban cemetery. According to her own request she was buried in Stanley Park, Vancouver's beautiful heritage of the forest primeval. A simple stone surrounded by rustic palings marks her grave and on this stone is carved the one word "Pauline." There she lies among ferns and wild flowers a short distance from Siwash Rock, the story of which she has recorded in the legends of her race. In time to come a pathway to her grave will be worn by lovers of Canadian poetry who will regard it as one of the most romantic of our literary shrines. (from the Biographical Sketch in the Flint and Feather collection)

25Erie Waters

By E. Pauline Johnson

"Her death is not only a great loss to those who knew and loved her: it is a great loss to Canadian literature and to the Canadian nation. I must think that she will hold a memorable place among poets in virtue of her descent and also in virtue of the work she has left behind, small as the quantity of that work is. I believe that Canada will, in future times, cherish her memory more and more, for of all Canadian poets she was the most distinctly a daughter of the soil, inasmuch as she inherited the blood of the great primeval race now so rapidly vanishing, and of the greater race that has supplanted it." (Theodore Watts-Dunton, from the Introduction to Flint and Feather)