This paper describes the Pedagogical Model of an Intelligent Tutoring System (ITS) for learning Computational Logic: SIAL. This ITS teaches classical refutation by resolution concepts. SIAL is a Model-Based System as its Domain Model is an Automated Theorem Prover (ATP) for First Order Logic (FOL). This allows SIAL to perform an accurate error diagnosis process while, at the same time, accepting several equivalent valid solutions to the same problem. This feature adds a general open-ended solution capability to SIAL which facilitates the adaptation to the students' way of thinking and allows them to explore and take their own decisions (and make their own mistakes). The Pedagogical Model of SIAL takes advantage of the error diagnosis capabilities of the Domain Model to offer a learner-adaptive tutorial action, according to the user cognitive profile. An example of the knowledge involved in the assessment of the next tutorial action based on the student model, and how SIAL adapts its behaviour to the student's assessed cognitive capabilities, is included. Some data obtained after a period of use of SIAL is also presented.

The binary classification problem where an input is classified as belonging or not to a certain class, the so-called Target Class (TC), is approached here. This problem can be stated as a basic hypothesis test: X is from the TC ( H 0 ) vs. X is not from the TC ( H 1 ), where X is the classifier input. When probabilistic models are used (e.g., Hidden Markov Models or Gaussian Mixture Models), the likelihood ratio, p ( X / H 0 )/ p ( X / H 1 ), is an alternative widely used to improve the classification. However, as far as we know, this ratio is not usually applied with distance-based classifiers (e.g., Dynamic Time Warping). Following that idea, here we propose making the decision based not only on the score (“score” being the classifier output) assuming X to be from the TC ( H 0 ), but also using the score assuming X is not from the TC ( H 1 ), by means of the ratio between both scores: the score ratio. The proposal is tested in biometric person authentication using manuscript signature, with three different state-of-the-art systems based on distance classifiers. Different alternatives for applying the proposal are shown in order to reduce the computer load, should it prove necessary. Using the score ratio has led to improvements in most of the tests performed. The best verification results were achieved using our proposal, with the best ones without the score ratio being improved by an average of 22%.

Score Normalization is a usual technique in pattern recognition to standardize the classifier output ranges so as to, for example, fuse these outputs. The use of score normalization in biometric recognition is a very important part of the system, specially in those based on behavioral traits, such as written signature or voice, conditioning the final system performance. Then, many works can be found that focus on the problem. A successful new approach for client threshold prediction, based on Multiple Linear Prediction, has been presented in recent works. Here, a new approach for score normalization, based on this proposal for biometric manuscript signature user verification, is shown. This proposal is compared with the state of the art methods, achieving an improvement of 19% and 16% for Equal Error Rate (EER) and 60% and 26% for Detection Cost Function (DCF) performance measures, for random and skilled forgeries, respectively.

This paper deals with the problem of training an Artificial Neural Network (ANN) when the data sets are very imbalanced. Most learning algorithms, including ANN, are designed for well-balanced data and do not work properly on imbalanced ones. Of the approaches proposed for dealing with this problem, we are interested in the re-sampling ones, since they are algorithm-independent. We have recently proposed a new under-sampling technique for the two-class problem, called Non-Target Incremental Learning (NTIL), which has shown a good performance with SVM, improving results and training speed. Here, the advantages of using this technique with ANN are shown. The performance with regard to other popular under-sampling techniques is compared.