On iterative learning from different tracking tasks in the presence of time-varying uncertainties.

In this paper, we introduce a new iterative learning control (ILC) method, which enables learning from different tracking control tasks. The proposed method overcomes the imitation of traditional ILC in that, the target trajectories of any two consecutive iterations can be completely different. For non-linear systems with time-varying and time-invariant parametric uncertainties, the new learning method works effectively to nullify the tracking error. To facilitate the learning control system design and analysis, in the paper we use a composite energy function (CEF) index, which consists of a positive scalar function and L2 norm of the function approximation error.     

Effects of user age and target-expansion methods on target-acquisition tasks using a mouse

Target expansion, i.e., the increase of target size according to cursor movement, can be a practical scheme to improve the usability of target-selection tasks using a mouse. This study examined the effects of different user age groups and target-expansion methods on target-acquisition tasks with grouped icons. Twenty-eight subjects performed acquisition tasks under eight experimental conditions: combinations of four expansion areas (no, one-icon, fish-eye, and group expansion) and two expansion techniques (occlusion and push). Older users took longer to acquire targets than younger users; however, they showed no significant difference in accuracy. Target expansion did not substantially improve performance speed compared to the static condition. However, the error rate was lowest when group area was expanded with the push technique, and both age groups were most satisfied with one-icon area expansion with the occlusion technique. We suggest alternative guidelines in designing target-expansion schemes.     

Autonomous learning of sequential tasks: experiments and analyses

Presents a learning model CLARION, which is a hybrid model based on the two-level approach proposed by Sun. The model integrates neural, reinforcement, and symbolic learning methods to perform on-line, bottom-up learning (i.e., learning that neural to symbolic representations). The model utilizes procedural and declarative knowledge (in neural and symbolic representations, respectively), tapping into the synergy of the two types of processes. It was applied to deal with sequential decision tasks. Experiments and analyses of various ways are reported that shed light on the advantages of the model.     

Accuracy-based learning classifier systems: models, analysis and applications to classification tasks

Recently, Learning Classifier Systems (LCS) and particularly XCS have arisen as promising methods for classification tasks and data mining. This paper investigates two models of accuracy-based learning classifier systems on different types of classification problems. Departing from XCS, we analyze the evolution of a complete action map as a knowledge representation. We propose an alternative, UCS, which evolves a best action map more efficiently. We also investigate how the fitness pressure guides the search towards accurate classifiers. While XCS bases fitness on a reinforcement learning scheme, UCS defines fitness from a supervised learning scheme. We find significant differences in how the fitness pressure leads towards accuracy, and suggest the use of a supervised approach specially for multi-class problems and problems with unbalanced classes. We also investigate the complexity factors which arise in each type of accuracy-based LCS. We provide a model on the learning complexity of LCS which is based on the representative examples given to the system. The results and observations are also extended to a set of real world classification problems, where accuracy-based LCS are shown to perform competitively with respect to other learning algorithms. The work presents an extended analysis of accuracy-based LCS, gives insight into the understanding of the LCS dynamics, and suggests open issues for further improvement of LCS on classification tasks.     

Individual and group-based learning from complex cognitive tasks: Effects on retention and transfer efficiency

The effects of individual versus group learning (in triads) on efficiency of retention and transfer test performance in the domain of biology (heredity) among 70 high-school students were investigated. Applying cognitive load theory, the limitations of the working memory capacity at the individual level were considered an important reason to assign complex learning tasks to groups rather than to individuals. It was hypothesized that groups will have more processing capacity available for relating the information elements to each other and by doing so for constructing higher quality cognitive schemata than individuals if the high cognitive load imposed by complex learning tasks could be shared among group members. In contrast, it was expected that individuals who learn from carrying out the same complex tasks would need all available processing capacity for remembering the interrelated information elements, and, consequently, would not be able to allocate resources to working with them. This interaction hypothesis was confirmed by the data on efficiency of retention and transfer test performance; there was a favorable relationship between mental effort and retention test performance for the individual learners as opposed to a favorable relationship between transfer test performance and mental effort for the students who learned in groups.     

Incremental learning of context free grammars based on bottom-up parsing and search

This paper describes approaches for machine learning of context free grammars (CFGs) from positive and negative sample strings, which are implemented in Synapse system. The grammatical inference consists of a rule generation by “inductive CYK algorithm,” mechanisms for incremental learning, and search. Inductive CYK algorithm generates minimum production rules required for parsing positive samples, when the bottom-up parsing by CYK algorithm does not succeed. The incremental learning is used not only for synthesizing grammars by giving the system positive strings in the order of their length but also for learning grammars from other similar grammars. Synapse can synthesize fundamental ambiguous and unambiguous CFGs including nontrivial grammars such as the set of strings not of the form ww with w∈{a,b}⁺.     

Incremental and accurate computation of machine learning models with smart data summarization

Journal of Intelligent Information Systems - Nowadays, data scientists prefer “easy” high-level languages like R and Python, which accomplish complex mathematical tasks with a few lines...     

Static and interactive infographics in daily tasks: A value-in-use and quality of interaction user study

Infographics are a common visual means to inform users. This paper investigates how lay people of different age, gender and educational background perceive the use of infographics for information visualization in daily tasks. We chose three topics of general interest: weather, study and work and three infographics, one for each topic. We administered a questionnaire to people randomly split in two groups: the first group interacted with a static version of each infographic, i.e., a snapshot of it; the second group interacted with the fully configurable infographics. We aimed to assess information quality on different dimensions, to take into account both formal and substantial aspects; interaction quality along dimensions like usability and ease of use; and design quality on the dimensions of the Visualization Wheel by Cairo, to assess the trade-off between information complexity and aesthetics of infographics. The goal was to measure whether the quality of infographics affects the perception of information and the users' interaction. The overall results suggest that, although interactive infographics are perceived as more complex, the experience with them is better. From our observations, we derived a model to assess the overall quality of static and interactive infographics, based on information, interaction and design quality dimensions.     

Incremental motion learning through kinesthetic teachings and new motion production from learned motions by a humanoid robot

This work presents a new incremental motion learning algorithm through kinesthetic teachings and a new motion production algorithm by combining learned motions in a humanoid robot. The proposed algorithms are useful for improving the motions that a humanoid robot can produce. The learning algorithm consists of data encoding, time alignment, dimensional reduction, parameter estimation in the Gaussian mixture model (GMM) of motions, GMM refinement, and motion generation steps. The overall procedure is built to be incremental. No historic data memorization is required in any step, and model parameters are enough information to generate motions. The motion production algorithm allows a robot to extract new motions simply from learned motions without requiring teaching sessions. A series of experiments with a humanoid robot serves to validate the performance of the proposed algorithms.     

An hybrid deep learning approach for depression prediction from user tweets using feature-rich CNN and bi-directional LSTM

Multimedia Tools and Applications - Depression has become one of the most widespread mental health disorders across the globe. Depression is a state of mind which affects how we think, feel, and...     

Effects of adapting to user pitch on rapport perception,behavior, and state with a social robotic learning companion

User Modeling and User-Adapted Interaction - Social robots such as learning companions, therapeutic assistants, and tour guides are dependent on the challenging task of establishing a rapport with...     

Page flipping vs. clicking: The impact of naturally mapped interaction technique on user learning and attitudes

Newer interaction techniques enable users to explore interfaces in a more natural and intuitive way. However, we do not yet have a scientific understanding of their contribution to user experience and theoretical mechanisms underlying the impact. This study examines how a naturally mapped interface, page-flipping interface, can influence user learning and attitudes. An online experiment with two conditions (page flipping vs. clicking) tests the impact of this naturally mapped interaction technique on user learning and attitudes. The result shows that the page-flipping feature creates more positive evaluations of the website in terms of usability and engagement, as well as greater behavioral intention towards the website by evoking greater perception of natural mapping and greater feeling of presence. In terms of learning outcomes, however, participants who flip through the online magazine show less recall and recognition memory, unless they perceive page flipping as more natural and intuitive to interact with. Participants perceive the same content as more credible when they flip through the content, but only if they appreciate the coolness of the medium. Theoretical and practical implications will be discussed.     

Predictability, complexity, and learning.

We define predictive information I(pred)(T) as the mutual information between the past and the future of a time series. Three qualitatively different behaviors are found in the limit of large observation times T:I(pred)(T) can remain finite, grow logarithmically, or grow as a fractional power law. If the time series allows us to learn a model with a finite number of parameters, then I(pred)(T) grows logarithmically with a coefficient that counts the dimensionality of the model space. In contrast, power-law growth is associated, for example, with the learning of infinite parameter (or nonparametric) models such as continuous functions with smoothness constraints. There are connections between the predictive information and measures of complexity that have been defined both in learning theory and the analysis of physical systems through statistical mechanics and dynamical systems theory. Furthermore, in the same way that entropy provides the unique measure of available information consistent with some simple and plausible conditions, we argue that the divergent part of I(pred)(T) provides the unique measure for the complexity of dynamics underlying a time series. Finally, we discuss how these ideas may be useful in problems in physics, statistics, and biology.     

Students’ expectations of,and experiences in e-learning: Their relation to learning achievements and course satisfaction

Within only a few years, the use of e-learning has increased rapidly in Austria. In certain subjects, up to 60% of university students report using e-learning platforms at least ‘sometimes’ or ‘frequently’ (Unger & Wroblewski, 2006). Yet, which aspects of e-learning do students consider important for their learning achievements and course satisfaction? This question was addressed by surveying 2196 students from 29 universities in Austria about their expectations of, and experiences in e-learning. Multiple regression analyses using Mplus 4.21 were carried out to investigate how different facets of students’ expectations and experiences are related to perceived learning achievements and course satisfaction.