Abstract: This article considers iterative learning control (ILC) for a class of discrete-time systems with full learnability and unknown system dynamics. First, we give a framework to analyze the learnability of the control system and build the relationship between the learnability of the control system and the input-output coupling matrix (IOCM). The control system has full learnability if and only if the IOCM is full-row rank and the control system has no learnability almost everywhere if and only if the rank of the IOCM is less than the dimension of system output. Second, by using the repetitiveness of the control system, some data-based learning schemes are developed. It is shown that we can obtain all the needed information on system dynamics through the developed learning schemes if the control system is controllable. Third, by the dynamic characteristics of system outputs of the ILC system along the iteration direction, we show how to use the available information of system dynamics to design the iterative learning gain matrix and the current state feedback gain matrix. And we strictly prove that the iterative learning scheme with the current state feedback mechanism can guarantee the monotone convergence of the ILC process if the IOCM is full-row rank. Finally, a numerical example is provided to validate the effectiveness of the proposed iterative learning scheme with the current state feedback mechanism.

Abstract: We study the sample complexity of learning revenue-optimal multi-item auctions. We obtain the first set of positive results that go beyond the standard but unrealistic setting of item-independence. In particular, we consider settings where bidders' valuations are drawn from correlated distributions that can be captured by Markov Random Fields or Bayesian Networks -- two of the most prominent graphical models. We establish parametrized sample complexity bounds for learning an up-to-e optimal mechanism in both models, which scale polynomially in the size of the model, i.e. the number of items and bidders, and only exponential in the natural complexity measure of the model, namely either the largest in-degree (for Bayesian Networks) or the size of the largest hyper-edge (for Markov Random Fields). We obtain our learnability results through a novel and modular framework that involves first proving a robustness theorem. We show that, given only "approximate distributions" for bidder valuations, we can learn a mechanism whose revenue is nearly optimal simultaneously for all "true distributions" that are close to the ones we were given in Prokhorov distance. Thus, to learn a good mechanism, it suffices to learn approximate distributions. When item values are independent, learning in Prokhorov distance is immediate, hence our framework directly implies the main result of Gonczarowski and Weinberg[36]. When item values are sampled from more general graphical models, we combine our robustness theorem with novel sample complexity results for learning Markov Random Fields or Bayesian Networks in Prokhorov distance, which may be of independent interest. Finally, in the single-item case, our robustness result can be strengthened to hold under an even weaker distribution distance, the Levy distance.

Abstract: In recent years we see a rapidly growing line of research which shows learnability of various models via common neural network algorithms. Yet, besides a very few outliers, these results show learnability of models that can be learned using linear methods. Namely, such results show that learning neural-networks with gradient-descent is competitive with learning a linear classifier on top of a data-independent representation of the examples. This leaves much to be desired, as neural networks are far more successful than linear methods. Furthermore, on the more conceptual level, linear models don't seem to capture the "deepness" of deep networks. In this paper we make a step towards showing leanability of models that are inherently non-linear. We show that under certain distributions, sparse parities are learnable via gradient decent on depth-two network. On the other hand, under the same distributions, these parities cannot be learned efficiently by linear methods.

Abstract: We prove that every concept class with finite Littlestone dimension can be learned by an (approximate) differentially-private algorithm This answers an open question of Alon et al (STOC 2019) who proved the converse statement (this question was also asked by Neel et al (FOCS 2019)) Together these two results yield an equivalence between online learnability and private PAC learnability We introduce a new notion of algorithmic stability called “global stability” which is essential to our proof and may be of independent interest We also discuss an application of our results to boosting the privacy and accuracy parameters of differentially-private learners

Abstract: The article addresses the issue of implementing the usability principles of educational internet resources. The paper debates the latest researches on the question concerning the search for the factors that influence the results of online education. The analysis, which we carried out, allowed us to focus on such known six criteria of usability design as Information Quality, System Navigation, System Learnability, Visual Design, Instructional Assessment, and System Interactivity and suggest the existence of the seventh criterion named Responsiveness. The research considers the principles of usability implementation following the example of the open platform of online education “Higher School Mathematics Teacher”. The answers given by 203 respondents during the survey allowed defining the direction of implementing the usability criteria on the platform. We were eager to know the opinion of teachers and students, who became the first users of the platform. The article discusses the criteria implementation while developing online courses on the platform. There was ground to conclude that when designing on-line platform courses, all seven usability subcategories are important.

Abstract: Most existing Business Process Model and Notation (BPMN) editing tools are graphical, and as such based on explicit modeling, requiring good knowledge of the notation and its semantics, as well as the ability to analyze and abstract business requirements and capture them by correctly using the notation. As a consequence, their use can be cumbersome for live modeling during interviews and design workshops, where participants should not only provide input but also give feedback on how it has been represented in a model. To overcome this, in this paper we present the design and evaluation of BPMN Sketch Miner, a tool which combines notes taking in constrained natural language with process mining to automatically produce BPMN diagrams in real-time as interview participants describe them with stories. In this work we discuss the design decisions regarding the trade-off between using mining vs. modelling in order to: 1) support a larger number of BPMN constructs in the textual language; 2) target both BPMN beginners and business analysts, in addition to the process participants themselves. The evaluation of the new version of the tool in terms of how it balances the expressiveness and learnability of its DSL with the usability of the text-to-visual sketching environment shows encouraging results. Namely, while BPMN beginners could model a non-trivial process with the tool in a relatively short time and with good accuracy, business analysts appreciated the usability of the tool and the expressiveness of the language in terms of supported BPMN constructs.

Abstract: This article addresses an important problem of how to improve the learnability of an intelligent agent in a strongly connected multiagent network. A novel spatial-dimensional linear dynamic relationship (SLDR) is developed to formulate the spatial dynamic relationship of an agent with respect to all the related agents. The obtained SLDR virtually exists in the computer to describe the input-output (I/O) relationship in the spatial domain and an iterative adaptation mechanism is developed to update the SLDR using I/O information to show real-time dynamical behavior of multiagent systems with nonrepetitive initial states. Subsequently, an SLDR-based adaptive iterative learning control (SLDR-AILC) is presented with rigorous analysis for iteration-variant formation control targets. Not only the 3-D dynamic behavior of the multiagent network but also the control protocols of the communicated agents are incorporated in the learning mechanism and thus strong learnability of the proposed SLDR-AILC is achieved to improve control performance. The proposed SLDR-AILC is a data-driven scheme where no explicit model structure is needed. Simulations with strongly connected topologies verify the theoretical results.

Abstract: Purpose This paper aims to explore the use of tangible user interfaces for teaching concepts related to internet of things by focusing on two aspects, notably, usability and learning effectiveness. Design/methodology/approach To assess the usability of IoTTT, Nielsen’s principles were used due to its relevance and popularity for usability assessment. In the usability questionnaire, four attributes were evaluated, notably, learnability, efficiency, errors and satisfaction. As for evaluating learning effectiveness, learning assessment was conducted through pre-tests and post-tests. Two groups of 20 students participated where the first group attended conventional lectures on IoT, whereas the second group used IoTTT for learning same concepts. In the process, data was collected through the usability questionnaire and tests for usability and learning effectiveness assessment. Findings Results revealed a positive score for the usability of the TUI solution with an average rating of 3.9. Although this score demonstrated an acceptable solution, different issues were identified, based on which a set of recommendations have been made in this paper. On the other hand, in the common pre-tests, an average score of 6.40 was obtained as compared to a mean score of 7.33 in the post-tests for all participants. Knowledge gains were significantly higher for students who learnt IoT concepts through the TUI-based system where performance improved by 18 per cent. Originality/value The results revealed in this study are expected to help the research community, course designers and tutors comprehend the prospects of using tangible user interfaces to foster teaching and learning of IoT concepts. In addition, educational solution providers could consider commercialisation prospects of this technology to innovate in teaching and learning, while also building-up on limitations identified within this study.

Abstract: This paper presents evidence that tone well-formedness patterns share a property of melody-locality, and shows how patterns with this property can be learned. Essentially, a melody-local pattern is one in which constraints over an autosegmental melody operate independently of constraints over the string of tone-bearing units. This includes a range of local tone patterns, long-distance tone-patterns, and their interactions. These results are obtained from the perspective of formal language theory and grammatical inference, which focus on the structural properties of patterns, but the implications extend to other learning frameworks. In particular, a melody-local learner can induce attested tone patterns that cannot be learned by the tier projection learners that have formed the basis of work on learning long-distance phonology. Thus, melody-local learning is a necessary property for learning tone. It is also shown how melody-local learners are more restrictive than learning directly over autosegmental representations.

Abstract: The study of strategic or adversarial manipulation of testing data to fool a classifier has attracted much recent attention. Most previous works have focused on two extreme situations where any testing data point either is completely adversarial or always equally prefers the positive label. In this paper, we generalize both of these through a unified framework for strategic classification, and introduce the notion of strategic VC-dimension (SVC) to capture the PAC-learnability in our general strategic setup. SVC provably generalizes the recent concept of adversarial VC-dimension (AVC) introduced by Cullina et al. arXiv:1806.01471. We instantiate our framework for the fundamental strategic linear classification problem. We fully characterize: (1) the statistical learnability of linear classifiers by pinning down its SVC; (2) its computational tractability by pinning down the complexity of the empirical risk minimization problem. Interestingly, the SVC of linear classifiers is always upper bounded by its standard VC-dimension. This characterization also strictly generalizes the AVC bound for linear classifiers in arXiv:1806.01471.

Abstract: Physically Unclonable Functions (PUFs) are widely adopted in various lightweight authenticating devices due to their unique fingerprints - providing uniform, unpredictable and reliable nature of responses. However, with the growth of machine learning (ML) attacks in recent times, it is imperative that the PUFs need to be resilient to such modeling attacks as well. Consequently, analyzing the learnability of PUFs has initiated a new branch of study leading to establishing provable guarantees (and PAC-learnability) of various PUF designs. However, these derivations are often carried out manually while implementing the design and thereby cannot automatically adjust the changes in PUF designs or its various compositions. In this paper, for the first time, we present an automated framework, called PUF-G, to reason about the PAC-learnability of PUF designs from an architectural level. To enable this, we propose a formal PUF representation language by which any architectural PUF design and its compositions can be specified upfront. This PUF specification can be automatically analyzed through a CAD framework by translating the same to an interim model and then deriving the PAC-learnability bounds from the model. Such a tool will help the designer to explore various compositional architectures of PUFs and its resilience to ML attacks automatically before converging on a strong PUF design for implementation. We also show the efficacy of our proposed framework over a wide range of PUF architectures while automatically deriving their learnability guarantees. As a matter of independent interest, the framework presents the first reported proofs to show that Interpose-PUF (newly proposed), MUX-PUF, FF-APUF, FF-XOR APUF and DA-PUF, are all PAC-learnable.

Abstract: Technology has become an essential part of the translation profession. Nowadays, computer-assisted translation (CAT) tools are extensively used by translators to enhance their productivity while maintaining high-quality translation services. CAT tools have gained popularity given that they provide a useful environment to facilitate and manage translation projects. Yet, little research has been conducted to investigate the usability of these tools, especially among Arab translators. In this study, we evaluate the usability of CAT tool from the translators’ perspective. The software usability measurement inventory (SUMI) survey is used to evaluate the system based on its efficiency, affect, usefulness, control, and learnability attributes. In total, 42 participants completed the online survey. Results indicated that the global usability of these tools is above the average. Results for all usability subscales were also above average wherein the highest scores were obtained for affect and efficiency, and the lowest scores were attributed to helpfulness and learnability. The findings suggest that CAT tool developers need to work further on the enhancement of the tool’s helpfulness and learnability to improve the translator’s experience and satisfaction levels. Further improvements are still required to increase the Arabic language support to meet the needs of Arab translators.

Abstract: We give a survey of the foundations of statistical queries and their many applications to other areas. We introduce the model, give the main definitions, and we explore the fundamental theory statistical queries and how how it connects to various notions of learnability. We also give a detailed summary of some of the applications of statistical queries to other areas, including to optimization, to evolvability, and to differential privacy.

Abstract: Mobile application is one of the most growth technology that can help people do better. Mobile applications nowadays have created tremendously compared with the website. Usability issues will be questioning as a measurement of mobile application accepted in the market. This paper reports on the systematic literature review (SLR) from 2010 up to 2020 to investigate how usability evaluation of mobile applications will change along the time through the evaluation of mobile devices. From 168 paper retrieved, 32 papers are selected to find the relevance of the research question. The result shows several famous indicators of usability evaluation, which are effectiveness, efficiency, and user satisfaction. There are new inclusions that come as new indicators such as error, learnability, memorability, simplicity, and cognitive load. This result indicates the need for future research especially for new rapid applications created

Abstract: This paper considers the problem of estimating the information leakage of a system in the black-box scenario, i.e. when the system's internals are unknown to the learner, or too complicated to analyze, and the only available information are pairs of input-output data samples, obtained by submitting queries to the system or provided by a third party. The frequentist approach relies on counting the frequencies to estimate the input-output conditional probabilities, however this method is not accurate when the domain of possible outputs is large. To overcome this difficulty, the estimation of the Bayes error of the ideal classifier was recently investigated using Machine Learning (ML) models, and it has been shown to be more accurate thanks to the ability of those models to learn the input-output correspondence. However, the Bayes vulnerability is only suitable to describe one-try attacks. A more general and flexible measure of leakage is the g-vulnerability, which encompasses several different types of adversaries, with different goals and capabilities. We propose a novel approach to perform black-box estimation of the g-vulnerability using ML which does not require to estimate the conditional probabilities and is suitable for a large class of ML algorithms. First, we formally show the learnability for all data distributions. Then, we evaluate the performance via various experiments using k-Nearest Neighbors and Neural Networks. Our approach outperform the frequentist one when the observables domain is large.

Abstract: We present refinements over existing temporal relation annotations in the Electronic Medical Record clinical narrative. We refined the THYME corpus annotations to more faithfully represent nuanced temporality and nuanced temporal-coreferential relations. The main contributions are in re-defining CONTAINS and OVERLAP relations into CONTAINS, CONTAINS-SUBEVENT, OVERLAP and NOTED-ON. We demonstrate that these refinements lead to substantial gains in learnability for state-of-the-art transformer models as compared to previously reported results on the original THYME corpus. We thus establish a baseline for the automatic extraction of these refined temporal relations. Although our study is done on clinical narrative, we believe it addresses far-reaching challenges that are corpus- and domain- agnostic.

Abstract: Alon et al. [2019] and Bun et al. [2020] recently showed that online learnability and private PAC learnability are equivalent in binary classification. We investigate whether this equivalence extends to multi-class classification and regression. First, we show that private learnability implies online learnability in both settings. Our extension involves studying a novel variant of the Littlestone dimension that depends on a tolerance parameter and on an appropriate generalization of the concept of threshold functions beyond binary classification. Second, we show that while online learnability continues to imply private learnability in multi-class classification, current proof techniques encounter significant hurdles in the regression setting. While the equivalence for regression remains open, we provide non-trivial sufficient conditions for an online learnable class to also be privately learnable.

Abstract: A supervised learning algorithm has access to a distribution of labeled examples, and needs to return a function (hypothesis) that correctly labels the examples The hypothesis of the learner is taken from some fixed class of functions (eg, linear classifiers, neural networks etc) A failure of the learning algorithm can occur due to two possible reasons: wrong choice of hypothesis class (hardness of approximation), or failure to find the best function within the hypothesis class (hardness of learning) Although both approximation and learnability are important for the success of the algorithm, they are typically studied separately In this work, we show a single hardness property that implies both hardness of approximation using linear classes and shallow networks, and hardness of learning using correlation queries and gradient-descent This allows us to obtain new results on hardness of approximation and learnability of parity functions, DNF formulas and $AC^0$ circuits

Abstract: Word Embeddings are used widely in multiple Natural Language Processing (NLP) applications. They are coordinates associated with each word in a dictionary, inferred from statistical properties of these words in a large corpus. In this paper we introduce the notion of "concept" as a list of words that have shared semantic content. We use this notion to analyse the learnability of certain concepts, defined as the capability of a classifier to recognise unseen members of a concept after training on a random subset of it. We first use this method to measure the learnability of concepts on pretrained word embeddings. We then develop a statistical analysis of concept learnability, based on hypothesis testing and ROC curves, in order to compare the relative merits of various embedding algorithms using a fixed corpora and hyper parameters. We find that all embedding methods capture the semantic content of those word lists, but fastText performs better than the others.

Abstract: The increasing impact of black box models, and particularly of unsupervised ones, comes with an increasing interest in tools to understand and interpret them. In this paper, we consider in particular how to characterise visual groupings discovered automatically by deep neural networks, starting with state-of-the-art clustering methods. In some cases, clusters readily correspond to an existing labelled dataset. However, often they do not, yet they still maintain an "intuitive interpretability". We introduce two concepts, visual learnability and describability, that can be used to quantify the interpretability of arbitrary image groupings, including unsupervised ones. The idea is to measure (1) how well humans can learn to reproduce a grouping by measuring their ability to generalise from a small set of visual examples (learnability) and (2) whether the set of visual examples can be replaced by a succinct, textual description (describability). By assessing human annotators as classifiers, we remove the subjective quality of existing evaluation metrics. For better scalability, we finally propose a class-level captioning system to generate descriptions for visual groupings automatically and compare it to human annotators using the describability metric.

Abstract: Person systems convey the roles entities play in the context of speech (e.g., speaker, addressee). Like other linguistic category systems, not all ways of partitioning the person space are equally ...

Abstract: People increasingly access cross-device applications from their smartphones while on the go. Yet, they do not fully use the mobile versions for complex tasks, preferring the desktop version of the same application. We conducted a survey (N=77) to identify challenges when switching back and forth between devices. We discovered significant cross-device learnability issues, including that users often find exploring the mobile version frustrating, which leads to prematurely giving up on using the mobile version. Based on the findings, we created four design concepts as video prototypes to explore how to support cross-device learnability. The concepts vary in four key dimensions: the device involved, automation, temporality, and learning approach. Interviews (N=20) probing the design concepts identified individual differences affecting cross-device learning preferences, and that users are more motivated to use cross-device applications when offered the right cross-device learnability support. We conclude with future design directions for supporting seamless cross-device learnability.

Abstract: Classification has been the prominent technique in machine learning domain, due to its ability of forecasting and predicts capabilities it is widely used in various domains such as health care, networking, social network, and software engineering with enhancement of different algorithm. The performance of the classifier majorly depends on the quality and amount of data present in the training sample. In real-world scenario, the majority of training samples suffered from class imbalance problem, that is, most of the data samples belong to one particular category, i.e., majority class while very few represent the minority class. In this case, classification techniques tend to be overwhelmed by the majority class and ignore the minority class. To solve class imbalance problem people relay on the different kind of sampling techniques either by generating synthetic data or by concentrating on minority class samples, but those approaches have introduced adverse effect in the learnability. In this paper, we attempt to study different techniques proposed to solve the class imbalance problem.

Abstract: Effort has been made to study the effect of medication-related clinical decision support systems in the inpatient setting; however, there is not much known about the usability of these systems. The goal of this study is to systematically review studies that focused on the usability aspects such as effectiveness, efficiency, and satisfaction of these systems. We systematically searched relevant articles in Scopus, Embase, and PubMed from 1 January 2000 to 1 January 2016, and found 22 articles. Based on Van Welie's usability model, we categorized usability aspects in terms of usage indicators and means. Our results showed that evidence was mainly found for effectiveness and efficiency. They showed positive results in the usage indicators errors and safety and performance speed. The means warnings and adaptability also had mostly positive results. To date, the effects satisfaction of clinical decision support system remains understudied. Aspects such as memorability, learnability, and consistency require more attention.

Abstract: Augmented Reality (AR) has become an emerging platform in the field of learning either in education or industry. Wearable AR smart glasses are the example of new wearable technology devices in AR. AR smart glasses has been introduced as another alternative to enhance the user’s experience in the real world rather than replacing the principle of learning. In education, AR smart glasses is seen as a possible technology to be embedded in the teaching and learning among lecturers and students for achieving smart campus environment. The effectiveness of AR smart glasses in education is a prospect for further study. Thus, this paper aims to identify the factors for investigating the learnability of students while learning using AR smart glasses. Various existing learning factors were reviewed through literature exploration and potential AR smart glasses learning factors are suggested. The results are significance for the development of the AR smart glasses learnability model that will further investigate in the future work.

Abstract: Given a quantum circuit, a quantum computer can sample the output distribution exponentially faster in the number of bits than classical computers. A similar exponential separation has yet to be established in generative models through quantum sample learning: given samples from an n-qubit computation, can we learn the underlying quantum distribution using models with training parameters that scale polynomial in n under a fixed training time? We study four kinds of generative models: Deep Boltzmann machine (DBM), Generative Adversarial Networks (GANs), Long Short-Term Memory (LSTM) and Autoregressive GAN, on learning quantum data set generated by deep random circuits. We demonstrate the leading performance of LSTM in learning quantum samples, and thus the autoregressive structure present in the underlying quantum distribution from random quantum circuits. Both numerical experiments and a theoretical proof in the case of the DBM show exponentially growing complexity of learning-agent parameters required for achieving a fixed accuracy as n increases. Finally, we establish a connection between learnability and the complexity of generative models by benchmarking learnability against different sets of samples drawn from probability distributions of variable degrees of complexities in their quantum and classical representations.

Abstract: Syntagmatic redundancy involves the multiple expressions of a single meaning within a phrase or clause. It is often claimed to be a linguistic universal that serves to facilitate expressivity, processing, and learnability. However, there is little empirical evidence supporting this theory. This paper combines a typological study of concord, a form of syntagmatic redundancy in which a lexical and a grammatical item with overlapping meanings are expressed in the same phrase or clause, with a functional analysis of concord. The purpose of the study was to find out if redundancy is indeed universal or whether there are cross-linguistic restrictions. The goal of the functional analysis was to provide better understanding of what motivates different forms of redundancy. Reference grammars of a 50-language variety sample were analyzed for the existence and communicative functions of four types of concord. The results show that argument concord and temporal concord are nearly universal, whereas only a subset of languages allow for negative concord and plural concord. Two functional principles are shown to motivate concord: the need to be precise, and the need to emphasize crucial information. These principles lead to distinct types of redundancy: The need to be precise results in accidental redundancy in the case of an obligatory grammatical marker, whereas the need to emphasize information invokes purposeful redundancy. The two types of redundancy are shown to be fundamentally distinct in their communicative nature as well as their characteristic diachronic development.

Abstract: In recent years we see a rapidly growing line of research which shows learnability of various models via common neural network algorithms. Yet, besides a very few outliers, these results show learnability of models that can be learned using linear methods. Namely, such results show that learning neural-networks with gradient-descent is competitive with learning a linear classifier on top of a data-independent representation of the examples. This leaves much to be desired, as neural networks are far more successful than linear methods. Furthermore, on the more conceptual level, linear models don't seem to capture the "deepness" of deep networks. In this paper we make a step towards showing leanability of models that are inherently non-linear. We show that under certain distributions, sparse parities are learnable via gradient decent on depth-two network. On the other hand, under the same distributions, these parities cannot be learned efficiently by linear methods.