Search

With the rapid advancement of Intelligent Transportation Systems (ITS) and vehicular communications, Vehicular Edge Computing (VEC) is emerging as a promising technology to support low-latency ITS applications and services. In this paper, we consider the computation offloading problem from mobile vehicles/users in a heterogeneous VEC scenario, and focus on the network- and base station selection problems, where different networks have different traffic loads. In a fast-varying vehicular environment, computation offloading experience of users is strongly affected by the latency due to the congestion at the edge computing servers co-located with the base stations. However, as a result of the non-stationary property of such an environment and also information shortage, predicting this congestion is an involved task. To address this challenge, we propose an on-line learning algorithm and an off-policy learning algorithm based on multi-armed bandit theory. To dynamically select the least congested network in a piece-wise stationary environment, these algorithms predict the latency that the offloaded tasks experience using the offloading history. In addition, to minimize the task loss due to the mobility of the vehicles, we develop a method for base station selection. Moreover, we propose a relaying mechanism for the selected network, which operates based on the sojourn time of the vehicles. Through intensive numerical analysis, we demonstrate that the proposed learning-based solutions adapt to the traffic changes of the network by selecting the least congested network, thereby reducing the latency of offloaded tasks. Moreover, we demonstrate that the proposed joint base station selection and the relaying mechanism minimize the task loss in a vehicular environment., Published in IEEE Transactions on Mobile Computing, Vol 21, Issue 12, Dec 2022

During Thailand’s COVID-19 lockdown, all university in the country must continue their classes in the Summer Semester with on-line format. This article described our university’s policy and practice to support university-wide on-line learning. Our support system included a Learning Management System, a teaching guideline for our instructors, an on-line training course for instructors, a quality control process, and a teaching assessment to obtain feedbacks from students. Due to the rushing nature of the situation, many instructors could not finish all teaching preparation before the semester started. However, the results showed that the students satisfied with their on-line classes with the average score of 4.54 from 5-point Likert Scale.

Covid-19 is (in) a class of its own in its influence on human lives and livelihoods globally, precipitating steep learning and psychological well-being curves for university teachers and students. This has impacted dramatically on the conditions under which higher education has had to function in regard to research and what is now referred to as ‘emergency online education’. As staff face unprecedented challenges, so too do students. Given that the consequences of these times are likely to be felt well into the future, it is important to capture what is happening now. We therefore present this perspective piece comprising 13 musings co-authored by students and educators regarding our experiences of two lockdowns within Aotearoa New Zealand in 2020 representative of the disrupted university in its adjustment to learning and teaching. In contributing to calls to develop a post-pandemic pedagogy for higher education, and better support staff and student well-being, we draw on methods that would multiply questions and invoke possibilities, as an impetus for reimagining higher education. Making use of a cogenerative dialoguing process, these musings enable multiple voices to be heard and considered. A non-representational lens enables us to explore the what and how of Covid-19 creating disruption and uncertainty for students’ and educators, influencing their psychological well-being and higher education pedagogy and practices, and becoming a contextually relevant taonga (treasure) of experiences that might inform future educational activities. © Copyright © 2021 Trafford, Haxell, Lau, Carlson, Rebelo da Silva, Hart, De Freitas and Rope.

Due to COVID-19, an annual interactive statistics and research methodology workshop for radiation oncology trainees was adapted at short notice into a live virtual format. This study aimed to evaluate trainee opinions around the educational value of the workshop, logistical aspects and impact on interactivity. A post-course on-line survey was completed by 26/42 trainee attendees (response rate 62%). For five pre-specified learning outcomes (LO), 58 to 69% of trainees agreed that the LO was completely or largely met (Likert scores 6 and 7 on a scale 1 = not met at all; 7 = completely met). All trainees felt that logistical aspects of the workshop including organisation, accessibility to the platform and sound/image quality were good or excellent. With regard to opportunities for interaction and suitability for small-group ‘break-out’ sessions, the majority felt that interaction could be adequately maintained whilst just under a quarter felt the delivery method was not fit for the purpose. Networking/social engagement with peers and teachers was the factor most impaired using the live virtual delivery format. Over three-quarters of trainees replied they would favour the current event or other educational sessions being offered (at least as an option) in a virtual format in the future. Cost and convenience were given as the major non-COVID-19–related benefits of virtual on-line learning. These preliminary findings provide valuable feedback to help adapt or develop further on-line educational and training initiatives that will be necessary in the COVID-19 pandemic period and beyond. Electronic supplementary material The online version of this article (10.1007/s13187-020-01898-9) contains supplementary material, which is available to authorized users.

The paper highlights the importance of monitoring students’ knowledge in the educational process organized on the basis of IT. The monitoring procedure in training enhances students’ knowledge. The results of this study show that the modernization of the monitoring in the educational process requires the use of modern computer equipment and means of communication. Electronic monitoring of students’ knowledge is the basis for obtaining an objective independent assessment of the level of students’ academic achievements (knowledge, intellectual and practical skills). The paper also presents the list of online services for creating broad-based tests, which include Google Forms, LEARNINGAPPS.ORG, PROPROFS, ClassMarker. Careful analysis of the learning forms allows us to present a way of organizing students’ training with the help of the Google platform, including the Google Classroom service (Learning Management System), which is essential for online learning and effective control of students’ knowledge of the subjects of their major. Unlike its predecessors, the Google Classroom as a separate learning management system allows teachers to use the integrated tools of the Google Apps package, such as Google Drive and Gmail. The main possibilities of using Google Classroom for the organization of Smart learning and monitoring of students’ knowledge are identified. The paper also presents detailed information on the possibilities for Google Classroom users (teachers, students, mentors, parents, administrators). The use of the Smart course, developed by means of the interactive services combined with the traditional methods of teaching information technologies, has made it possible to move the communication of students and teachers up to a new level – it is not limited only to a classroom. In the Google Classroom the educators can post tasks, send announcements, start discussions, and students can share materials, add comments on the course, communicate via e-mail. The information on the work done is constantly updated, which allows the educators to check the tasks quickly and monitor students’ knowledge. The article proves the effectiveness of using the Smart course on the basis of Google Classroom for the formation of information skills of students majoring in “Information and Software Processing Operator”.

A new online multi-class learning algorithm is proposed with three main characteristics. First, in order to make the feature pool fitter for the pattern pool, the adaptive feature pool is proposed to dynamically combine the three general features, Haar-like, Histogram of Oriented Gradient (HOG), and Local Binary Patterns (LBP). Second, the external model is integrated into the proposed model without re-training to enhance the efficacy of the model. Third, a new multi-class learning and updating mechanism are proposed that help to find unsuitable decisions and adjust them automatically. The performance of the proposed model is validated with multi-class detection and online learning system. The proposed model achieves a better score than other non-deep learning algorithms used in public pedestrian and multi-class databases. The multi-class databases contain data for pedestrians, faces, vehicles, motorcycles, bicycles, and aircraft.

Many educational institutes are recording classroom lessons to improve the students' learning process. Unfortunately, a simple recording does not necessarily imply benefits. Indeed, when producing video lectures, there are different challenges that must be faced: production costs, accessibility, usability, video indexing, just to name a few. In this paper, we share our experience building ONELab, a system designed to capture, record, edit and stream video lectures. ONELab was designed to be scalable and to have low-cost implementation and maintenance. The system has been used in the 2017-18 Academic Year to manage the 49 courses offered by the five degrees available at our Department. In numbers, it supported 1,054 freshman students and produced 1,376 video lectures (for a total of 2,064 hours). The usage analysis showed that students appreciated the system and a comparative analysis between students who used the system and students who did not use the system, showed that the former passed more exams (+97.8%), had better grades (+8%) and acquired more credits (+105%).

We introduce a modeling framework for the investigation of on-line machine learning processes in non-stationary environments. We exemplify the approach in terms of two specific model situations: In the first, we consider the learning of a classification scheme from clustered data by means of prototype-based Learning Vector Quantization (LVQ). In the second, we study the training of layered neural networks with sigmoidal activations for the purpose of regression. In both cases, the target, i.e., the classification or regression scheme, is considered to change continuously while the system is trained from a stream of labeled data. We extend and apply methods borrowed from statistical physics which have been used frequently for the exact description of training dynamics in stationary environments. Extensions of the approach allow for the computation of typical learning curves in the presence of concept drift in a variety of model situations. First results are presented and discussed for stochastic drift processes in classification and regression problems. They indicate that LVQ is capable of tracking a classification scheme under drift to a non-trivial extent. Furthermore, we show that concept drift can cause the persistence of sub-optimal plateau states in gradient based training of layered neural networks for regression.

Embedded, continual learning for autonomous and adaptive behavior is a key application of neuromorphic hardware. However, neuromorphic implementations of embedded learning at large scales that are both flexible and efficient have been hindered by a lack of a suitable algorithmic framework. As a result, the most neuromorphic hardware is trained off-line on large clusters of dedicated processors or GPUs and transferred post hoc to the device. We address this by introducing the neural and synaptic array transceiver (NSAT), a neuromorphic computational framework facilitating flexible and efficient embedded learning by matching algorithmic requirements and neural and synaptic dynamics. NSAT supports event-driven supervised, unsupervised and reinforcement learning algorithms including deep learning. We demonstrate the NSAT in a wide range of tasks, including the simulation of Mihalas-Niebur neuron, dynamic neural fields, event-driven random back-propagation for event-based deep learning, event-based contrastive divergence for unsupervised learning, and voltage-based learning rules for sequence learning. We anticipate that this contribution will establish the foundation for a new generation of devices enabling adaptive mobile systems, wearable devices, and robots with data-driven autonomy.

© 2018 The Authors. Published by Elsevier Ltd. This paper proposes a method to allow users to select target species for detection, generate an initial detection model by selecting a small piece of image sample and as the movie plays, continue training this detection model automatically. This method has noticeable detection results for several types of objects. The framework of this study is divided into two parts: the initial detection model and the online learning section. The detection model initialization phase use a sample size based on the proportion of users of the Haar-like features to generate a pool of features, which is used to train and select effective classifiers. Then, as the movie plays, the detecting model detects the new sample using the NN Classifier with positive and negative samples and the similarity model calculates new samples based on the fusion background model to calculate a new sample and detect the relative similarity to the target. From this relative similarity-based conservative classification of new samples, the conserved positive and negative samples classified by the video player are used for automatic online learning and training to continuously update the classifier. In this paper, the results of the test for different types of objects show the ability to detect the target by choosing a small number of samples and performing automatic online learning, effectively reducing the manpower needed to collect a large number of image samples and a large amount of time for training. The Experimental results also reveal good detection capability.

In this paper, a new algorithm for on-line learning of Random Vector Functional-Link neural network is proposed. It is specifically tailored to hardware implementations with finite precision arithmetic in distributed computing scenarios, where the massive use of low-cost hardware resources for sensor networks or computing agents is necessary in order to deal with big data, IoT paradigms and multiple sources of information. The proposed algorithm does not require any specific DSP operation to be implemented in hardware, like matrix inversions or multiplications, for real-time learning. However, experimental results prove that the algorithm outperforms commonly adopted recursive least-squares algorithms optimized for hardware implementation, while the loss of performance with respect to batch training is reduced, as the proposed approach is proved to be very efficient when a reduced number of bits is adopted for finite precision hardware implementation.

Bidirectional brain-machine interfaces (BMIs) establish a two-way direct communication link between the brain and the external world. A decoder translates recorded neural activity into motor commands and an encoder delivers sensory information collected from the environment directly to the brain creating a closed-loop system. These two modules are typically integrated in bulky external devices. However, the clinical support of patients with severe motor and sensory deficits requires compact, low-power, and fully implantable systems that can decode neural signals to control external devices. As a first step toward this goal, we developed a modular bidirectional BMI setup that uses a compact neuromorphic processor as a decoder. On this chip we implemented a network of spiking neurons built using its ultra-low-power mixed-signal analog/digital circuits. On-chip on-line spike-timing-dependent plasticity synapse circuits enabled the network to learn to decode neural signals recorded from the brain into motor outputs controlling the movements of an external device. The modularity of the BMI allowed us to tune the individual components of the setup without modifying the whole system. In this paper, we present the features of this modular BMI and describe how we configured the network of spiking neuron circuits to implement the decoder and to coordinate it with the encoder in an experimental BMI paradigm that connects bidirectionally the brain of an anesthetized rat with an external object. We show that the chip learned the decoding task correctly, allowing the interfaced brain to control the object's trajectories robustly. Based on our demonstration, we propose that neuromorphic technology is mature enough for the development of BMI modules that are sufficiently low-power and compact, while being highly computationally powerful and adaptive., Frontiers in Neuroscience, 10, ISSN:1662-453X, ISSN:1662-4548

Developing a robust and sustainable system is an important problem in which deep learning models are used in real-world applications. Ensemble methods combine diverse models to improve performance and achieve robustness. The analysis of time series data requires dealing with continuously incoming instances, however, most ensemble models suffer when adapting to a change in data distribution. Therefore, we propose an on-line ensemble deep learning algorithm that aggregates deep learning models and adjusts the ensemble weight based on loss value in this study. We theoretically demonstrate that the ensemble weight converges to the limiting distribution, and, thus, minimizes the average total loss from a new regret measure based on adversarial assumption. We also present an overall framework that can be applied to analyze time series. In the experiments, we focused on the on-line phase, in which the ensemble models predict the binary class for the simulated data and the financial and non-financial real data. The proposed method outperformed other ensemble approaches. Moreover, our method was not only robust to the intentional attacks but also sustainable in data distribution changes. In the future, our algorithm can be extended to regression and multiclass classification problems.

Incremental algorithms for fuzzy classifiers are studied in this paper. It is assumed that not all training patterns are given a priori for training classifiers, but are gradually made available over time. It is also assumed that the previously available training patterns can not be used afterwards. Thus, fuzzy classifiers should be modified by updating already constructed classifiers using the available training patterns. In this paper, a confidence-weighted (CW) learning algorithm is applied to fuzzy classifiers for this task. A series of computational experiments are conducted in order to examine the performance of the proposed method comparing that method with the conventional learning algorithm for fuzzy classifiers.

Date of Conference: 16-19 May 2016 Conference Name: IEEE 24th Signal Processing and Communications Applications Conference, SIU 2016 We introduce an on-line classification algorithm based on the hierarchical partitioning of the feature space which provides a powerful performance under the defined empirical loss. The algorithm adaptively partitions the feature space and at each region trains a different classifier. As a final classification result algorithm adaptively combines the outputs of these basic models which enables it to create a linear piecewise classifier model that can work well under highly non-linear complex data. The introduced algorithm also have scalable computational complexity that scales linearly with dimension of the feature space, depth of the partitioning and number of processed data. Through experiments we show that the introduced algorithm outperforms the state-of-the-art ensemble techniques over various well-known machine learning data sets.

By and large, the interpretation of learning as a computational process taking place in both humans and machines is primarily provided in the framework of statistics. In this paper, we propose a radically different perspective in which the emergence of learning is regarded as the outcome of laws of nature that govern the interactions of intelligent agents with their own environment. We introduce a natural learning theory based on the principle of least cognitive action, which is inspired to the related mechanical principle, and to the Hamiltonian framework for modeling the motion of particles. The introduction of the kinetic and of the potential energy leads to a surprisingly natural interpretation of learning as a dissipative process. The kinetic energy reflects the temporal variation of the synaptic connections, while the potential energy is a penalty that describes the degree of satisfaction of the environmental constraints. The theory gives a picture of learning in terms of the energy balancing mechanisms, where the novel notions of boundary and bartering energies are introduced. Finally, as an example of application of the theory, we show that the supervised machine learning scheme can be framed in the proposed theory and, in particular, we show that the Euler-Lagrange differential equations of learning collapse to the classic gradient algorithm on the supervised pairs.

Background: Supportive environments contribute to the mental health of individuals, local and global communities. Ways of thinking “critically” about what we know and ways of accomplishing mental health promotion work have a more contested history however. A critical perspective demands subjective awareness, and an openness to view the world from multiple perspectives. As such, a collaborative, international, and fully on-line course was developed between post-secondary educational institutions in Norway and Canada. This course intends to meet a need for critical awareness and understanding of the connection of diverse perspectives and environments to mental health promotion practices in the field. Purpose: Our study, of the development and implementation of an innovative critical mental health promotion course, was oriented around the following key questions: How is capacity for critical mental health promotion awareness developed through an on-line international learning environment? And how can supportive learning environments reshape mental health support across international borders? Our goal was to explore the work of students and faculty with the tensions that dominate current mental health promotion. This course began from the diverse locations, experiences and approaches of our students as important sites of critical exploration. Through the virtual world of electronic learning activities, students entered this diversity of worlds, and debated over the shape of health and mental health as notions themselves, as well as over the approaches that best produce support. Most importantly, students debated over the distribution of power, justice, and social and economic advantage held within mental health and various approaches to promoting mental health. Method: The overall approach was based on action research using collaborative enquiry of the case study of our pilot course offering. The enquiry was conducted over 3 years of course development and pilot course delivery. Results: Ultimately, through our reflection and study of the critical learning approaches in action, we were able to explore what supportive environments for mental health are and what particular approaches “do” locally and globally. Further we were able to reorient mental health promotion work through network building in ways that may potentially change nursing practice. Conclusions: The results and lessons learned, from development to implementation of an on-line collaborative international mental health promotion course, expanded the capacities of nursing programs, educational institutions, faculty and students, and expanded what can be known about supportive environments for mental health.

Highlights ► Classifier grids showed excellent detection results for stationary cameras. ► On-line adaptive classifiers reduce the complexity of the classification task. ► Fixed update strategies allow long-term stability. ► Short-term stability through proposed inverse multiple instance learning., Classifier grids have shown to be a considerable choice for object detection from static cameras. By applying a single classifier per image location the classifier’s complexity can be reduced and more specific and thus more accurate classifiers can be estimated. In addition, by using an on-line learner a highly adaptive but stable detection system can be obtained. Even though long-term stability has been demonstrated such systems still suffer from short-term drifting if an object is not moving over a long period of time. The goal of this work is to overcome this problem and thus to increase the recall while preserving the accuracy. In particular, we adapt ideas from multiple instance learning (MIL) for on-line boosting. In contrast to standard MIL approaches, which assume an ambiguity on the positive samples, we apply this concept to the negative samples: inverse multiple instance learning. By introducing temporal bags consisting of background images operating on different time scales, we can ensure that each bag contains at least one sample having a negative label, providing the theoretical requirements. The experimental results demonstrate superior classification results in presence of non-moving objects.

Making material available through learning management systems is standard practice in most universities, but this is generally seen as an adjunct to the ‘real’ teaching, that takes place in face-to-face classes. Lecture attendance is poor, and it is becoming increasingly difficult to engage students, both in the material being taught and campus life. This paper describes the redevelopment of a large course in scientific practice and communication that is compulsory for all science students studying at our Melbourne and Malaysian campuses, or by distance education. Working with an educational designer, a blended learning methodology was developed, converting the environment provided by the learning management system into a teaching space, rather than a filing system. To ensure focus, topics are clustered into themes with a ‘question of the week’, a pre-class stimulus and follow up activities. The content of the course did not change, but by restructuring the delivery using educationally relevant design techniques, the content was contextualised resulting in an integrated learning experience. Students are more engaged intellectually, and lecture attendance has improved. The approach we describe here is a simple and effective approach to bringing this university’s teaching and learning into the 21st century.

In this paper, a video based algorithm for fire and flame detection is developed. In addition to ordinary motion and color clues, flame flicker is distinguished from motion of flame colored moving objects using Markov models. Irregular nature of flame boundaries is detected by performing temporal wavelet analysis using Hidden Markov Models as well. Color variations in fire is detected by computing the spatial wavelet transform of moving fire-colored regions. Boundary of flames are represented in wavelet domain and irregular nature of the boundaries of fire regions is also used as an indication of the flame flicker. Decisions from sub-algorithms are linearly combined using an adaptive active fusion method. The main detection algorithm is composed of four sub-algorithms (i) detection of fire colored moving objects, (ii) temporal, and (iii) spatial wavelet analysis for flicker detection and (iv) contour analysis of fire colored region boundaries. Each algorithm yields a continuous decision value as a real number in the range [-1, 1] at every image frame of a video sequence. Decision values from sub-algorithms are fused using an adaptive algorithm in which weights are updated using the least mean square (LMS) method in the training (learning) stage.

There has been tremendous growth both in study abroad and intercultural exchange options in the United States and also in on-line learning options in higher education. Reviewing 91 cross-cultural experiences with at least one type of formal on-line component, this paper provides an overview and categorization of these offerings as well as a discussion of trends. It also offers recommendations to help strengthen and shape further offerings and related research.

Learning from the behavior of the resident is essential in order to adapt the lightning system and to provide intelligent lighting control based on behavior patterns. Different homes have different conditions and habits which have to be taken into account for the intelligent system to be useful. However, with passage of time even deeply ingrained habits are subject to change. Therefore, a truly intelligent system has to respond to the changing and diverse environment. An intelligent lighting control system that employs artificial neural networks for on-line learning and adaptation and is based on resident’s behavioural patterns is presented in this paper. In order to manage problems related to constant increase of data and dynamic environments, a group of algorithms have been improved by implementing a similarity threshold based data replacement algorithm that has been experimentally tested and compared with alternative algorithms.

Winner-Takes-All (WTA) prescriptions for learning vector quantization (LVQ) are studied in the framework of a model situation: two competing prototype vectors are updated according to a sequence of example data drawn from a mixture of Gaussialls. The theory of on-line learning allows for an exact mathernatical description of the training dynamics, even if an underlying cost function cannot be identified. We compare the typical behavior of several WTA schemes including basic LVQ and unsupervised vector quantization. The focus is on the learning curves, i.e. the achievable generalization ability as a function of the number of training examples. (c) 2005 Elsevier B.V. All rights reserved.

This article presents a plastic injection moulding monitoring module based on knowledge built on-line using feedback from production data. A fuzzy classifier was especially developed for this application. It is based on unsupervised and supervised classification methods. The role of the first one is to identify the functioning modes of the process whereas the role of the second one is to associate the state of the process to one of the identified functioning modes at the moment where a workpiece is injected. Furthermore this diagnosis module integrates an on-line learning method which allows to enrich and upgrade the initial knowledge during production. The results obtained show that the monitoring system is a solution for quality and productivity control having serious economical advantages. For example maintenance tasks can be anticipated and the size of the training set can be considerably reduced. The computing times show that the monitoring system can be used for the purpose of industrial applications without any decrease of production rate.

[EN] This paper proposes a multiple kernel regression (MKr) to predict water demand in the presence of a continuous source of infor- mation. MKr extends the simple support vector regression (SVR) to a combination of kernels from as many distinct types as kinds of input data are available. In addition, two on-line learning methods to obtain real time predictions as new data arrives to the system are tested by a real-world case study. The accuracy and computational efficiency of the results indicate that our proposal is a suitable tool for making adequate management decisions in the smart cities environment., This work has been supported by project IDAWAS, DPI2009- 11591, of the Direccion General de Investigacion of the Ministerio de Ciencia e Innovacion of Spain, and ACOMP/ 2011/ 188 of the Conselleria d'Educacio of the Generalitat Valenciana.

We propose a manifold regularization algorithm designed to work in an on-line scenario where data arrive continuously over time and it is not feasible to completely store the data stream for training the classifier in batch mode. The On-line Laplacian One-Class SVM (OLapOCSVM) algorithm exploits both positively labeled and totally unlabeled examples, updating the classifier hypothesis as new data becomes available. The learning procedure is based on conjugate gradient descent in the primal formulation of the SVM. The on-line algorithm uses an efficient buffering technique to deal with the continuous incoming data. In particular, we define a buffering policy that is based on the current estimate of the support of the input data distribution. The experimental results on real-world data show that OLapOCSVM compares favorably with the corresponding batch algorithms, while making it possible to be applied in generic on-line scenarios with limited memory requirements.

Multiview learning has been shown to be a natural and efficient framework for supervised or semi-supervised learning of multilingual document categorizers. The state-of-the-art co-regularization approach relies on alternate minimizations of a combination of language-specific categorization errors and a disagreement between the outputs of the monolingual text categorizers. This is typically solved by repeatedly training categorizers on each language with the appropriate regularizer. We extend and improve this approach by introducing an on-line learning scheme, where language-specific updates are interleaved in order to iteratively optimize the global cost in one pass. Our experimental results show that this produces similar performance as the batch approach, at a fraction of the computational cost., SIGIR 2012 The 35th International ACM SIGIR conference on research and development in Information Retrieval, August 12-16, 2012, Portland, Oregon, USA

Background The Public Health (PH) course at the medical college of Cairo University is based on traditional lectures. Large enrollment limits students' discussions and interactions with instructors. Aim Evaluate students' learning outcomes as measured by improved knowledge acquisition and opinions of redesigning the Reproductive Health (RH) section of the PH course into e-learning and assessing e-course utilization. Methods This prospective interventional study started with development of an e-learning course covering the RH section, with visual and interactive emphasis, to satisfy students' diverse learning styles. Two student groups participated in this study. The first group received traditional lecturing, while the second volunteered to enroll in the e-learning course, taking online course quizzes. Both groups answered knowledge and course evaluation questionnaires and were invited to group discussions. Additionally, the first group answered another questionnaire about reasons for non-participation. Results Students participating in the e-learning course showed significantly better results, than those receiving traditional tutoring. Students who originally shunned the e-course expressed eagerness to access the course before the end of the academic year. Overall, students using the redesigned e-course reported better learning experiences. Conclusions An online course with interactivities and interaction, can overcome many educational drawbacks of large enrolment classes, enhance student's learning and complement pit-falls of large enrollment traditional tutoring.

In this paper, a novel on-line regression method is presented. Due to its origins in Adaptive Resonance Theory neural networks, this method is particularly well-suited to problems requiring stable incremental learning. Its performance on five publicly available datasets is shown to be at least comparable to two established off-line methods. Furthermore, it exhibits considerable improvements in comparison to its closest supervised relative Fuzzy ARTMAP.

In on-line machine learning, predicting changes is not a trivial task. In this paper, a novel prediction approach is presented, that relies on a committee of experts. Each expert is trained on a specific history of changes and tries to predict future changes. The experts are constantly modified based on their performance and the committee as a whole is thus dynamic and can adapt to a large variety of changes. Experimental results based on synthetic data show three advantages: (a) it can adapt to different types of changes, (b) it can use different types of prediction models and (c) the committee outperforms predictors trained on a priori fixed size history of changes.

New variations on the application of the passive-aggressive algorithm to statistical machine translation are developed and compared to previously existing approaches. In online adaptation, the system needs to adapt to real-world changing scenarios, where training and tuning only take place when the system is set-up for the first time. Post-edit information, as described by a given quality measure, is used as valuable feedback within the passive-aggressive framework, adapting the statistical models on-line. First, by modifying the translation model parameters, and alternatively, by adapting the scaling factors present in stateof- the-art SMT systems. Experimental results show improvements in translation quality by allowing the system to learn on a sentence-by-sentence basis., This paper is based upon work supported by the EC (FEDER/FSE) and the Spanish MICINN under projects MIPRCV “Consolider Ingenio 2010” (CSD2007-00018) and iTrans2 (TIN2009-14511). Also supported by the Spanish MITyC under the erudito.com (TSI-020110-2009-439) project, by the Generalitat Valenciana under grant Prometeo/2009/014 and scholarship GV/2010/067 and by the UPV under grant 20091027.

In this paper, a novel unsupervised neural network combining elements from Adaptive Resonance Theory and topology learning neural networks, in particular the Self-Organising Incremental Neural Network, is introduced. It enables stable on-line clustering of stationary and non-stationary input data. In addition, two representations reflecting different levels of detail are learnt simultaneously. Furthermore, the network is designed in such a way that its sensitivity to noise is diminished, which renders it suitable for the application to real-world problems.

Cataloged from PDF version of article. There has been an increasing interest in the study of video based fire detection algorithms as video based surveillance systems become widely available for indoor and outdoor monitoring applications. A novel method explicitly developed for video based detection of wildfires at night (in the dark) is presented in this paper. The method comprises four sub-algorithms: (i) slow moving video object detection, (ii) bright region detection, (iii) detection of objects exhibiting periodic motion, and (iv) a sub-algorithm interpreting the motion of moving regions in video. Each of these sub-algorithms characterizes an aspect of fire captured at night by a visible range PTZ camera. Individual decisions of the sub-algorithms are combined together using a least-mean-square (LMS) based decision fusion approach, and fire/nofire decision is reached by an active learning method. (C) 2009 Elsevier Ltd. All rights reserved

We present a background subtraction approach aimed at efficiency and robustness to common sources of disturbance such as illumination changes, camera gain and exposure variations, noise. The novelty relies in trying to learn, at each new frame, a model of the background intensity changes currently yielded by the sources of disturbance. Based on the observation that such changes are highly correlated across large portions of the image, a unique frame-wise model is used, which consists in the bivariate probability density function of background and frame intensity at a pixel. After the non-parametric estimation of the model through the bivariate histogram, changes are detected by thresholding the histogram entries. Experimental results prove that the approach is state-of-the-art in challenging sequences characterized by sources of disturbance yielding sudden and strong background appearance changes. (6 pages)

International audience; In this paper, we describe how a mobile robot under simple visual control can retrieve a particular goal location in an open environment. Our model neither needs a precise map nor to learn all the possible positions in the environment. The system is a neural architecture inspired by neurobiological analysis of how visual patterns named landmarks are recognized. The robot merges these visual informations and their azimuth to build a plastic representation of its location. This representation is used to learn the best movement to reach the goal. A simple and fast on-line learning of a few places located near the goal allows this goal to be reached from anywhere in its neighborhood. The system uses only a very rough representation of the robot environment and presents very high generalization capabilities. We describe an efficient implementation of autonomous and motivated navigation tested on our robot in real indoor environments. We show the limitations of the model and its possible extensions.

International Conference on Artificial Neural Networks (IWANN), 1991, Granada (España), We have undertaken the study of methods for avoiding catastrophic forgetting in feedforward neural networks, without sacrifying the benefits of distributed representations. We formalize the problem as the minimization of the error over the previously learned input-output (i–o) patterns, subject to the constraint of perfect encoding of the new pattern. Then we transform this constrained optimization problem into an unconstrained one. This new formulation naturally leads to an algorithm for solving the problem, wihch we call Minimally Disturbing Learning (MDL). Some experimental comparisons of the performance of MDL with back-propagation are provided which, besides showing the advantages of using MDL, reveal the dependence of forgetting on the learning rate in back-propagation.

Objective Blended learning is a relatively new technology-based teaching approach. Few attempts have been made to use this approach in medical education. The aim of this study was to assess the effectiveness of blended learning in studying family medicine as an example of a clinical medical science. Methods This comparative cross-sectional study involved 121 fourth-year medical students, at the clinical phase of a family medicine course at the College of Medicine, Taibah University, Almadinah Almunawwarah, Kingdom of Saudi Arabia. The students were randomly divided into two groups: 61 taught by the traditional approach (face-to-face) and 60 taught by blended learning (both electronic and face-to-face). The effect of blended learning was evaluated from responses to the Dundee ‘ready education environment measure’ questionnaire. Clinical skills were assessed with the ‘objective structured clinical examination’; knowledge gain was assessed from a written multiple-choice examination; and problem-solving, critical thinking, decision-making skills and attitude were assessed in written and oral examinations based on clinical scenarios. Results Blended learning was statistically significantly better than traditional learning in all domains of the educational environment, except for social perception, and in all types of examination: written, objective structured clinical and case scenarios. Conclusion The findings suggest that medical students, as adult learners, are open to new methods of learning. The blended learning approach is an effective method for teaching family medicine and may be applicable to other clinical medical sciences.

We survey a number of recent results concerning the behaviour of algorithms for learning classifiers based on the solution of a regularized least-squares problem.

While there has been a significant amount of theoretical and empirical research on the multiple-instance learning model, most of this research is for concept learning. However, for the important application area of drug discovery, a real-valued classification is preferable. In this paper we initiate a theoretical study of real-valued multiple-instance learning. We prove that the problem of finding a target point consistent with a set of labeled multiple-instance examples (or bags) is NP-complete, and that the problem of learning from real-valued multiple-instance examples is as hard as learning DNF. Another contribution of our work is in defining and studying a multiple-instance membership query (MI-MQ). We give a positive result on exactly learning the target point for a multiple-instance problem in which the learner is provided with a MI-MQ oracle and a single adversarially selected bag.

We consider the problem of learning to predict as well as the best in a group of experts making continuous predictions. We assume the learning algorithm has prior knowledge of the maximum number of mistakes of the best expert. We propose a new master strategy that achieves the best known performance for on-line learning with continuous experts in the mistake bounded model. Our ideas are based on drifting games, a generalization of boosting and on-line learning algorithms. We prove new lower bounds based on the drifting games framework which, though not as tight as previous bounds, have simpler proofs and do not require an enormous number of experts. We also extend previous lower bounds to show that our upper bounds are exactly tight for sufficiently many experts. A surprising consequence of our work is that continuous experts are only as powerful as experts making binary or no prediction in each round.