We develop a method that is based on processing gathered Event Related Potentials (ERP) signals and the use of machine learning technique for multivariate analysis (i.e. classification) that we apply in order to analyze the differences between Dyslexic and Skilled readers.No human intervention is needed in the analysis process. This is the state of the art results for automatic identification of Dyslexic readers using a Lexical Decision Task. We use mathematical and machine learning based techniques to automatically discover novel complex features that (i) allow for reliable distinction between Dyslexic and Normal Control Skilled readers and (ii) to validate the assumption that most of the differences between Dyslexic and Skilled readers are located in the left hemisphere.Interestingly, these tools also pointed to the fact that High Pass signals (typically considered as "noise" during ERP/EEG analyses) in fact contain significant relevant information.Finally, the proposed scheme can be used for analysis of any ERP based studies.

In recent work, attacks on automated classification of Parkinson's disease have encountered difficulties, especially for cross-individual generalization. This is crucial since (i) Classifying the degree of Parkinson's disease is an important clinical necessity. (ii) The lack of such an automated system leaves current clinical methodology to use manual and subjective classification by a trained clinician. In earlier work, two of the authors of this paper have shown that, directly from the speech signal, reliable classification as to the presence of the disease can be produced using a machine learning approach. However, this approach was unable to reliably classify the severity degree of the disease. In other work, a deep (convolutional) neural network was tried on the same data set (albeit without feature extraction), which again did not succeed on the multi-label case. In this work, we applied a data science approach to solve this problem by analysing the topological structure of the label space and the internal topological structure of the data. Specifically we explored using (i) the linearity of the label-space to reduce the inherent noise in multi-class classifiers and (ii) to break the data into separate topological clusters (by using a version of unsupervised topological learning) and then applying separate classification parametrizations for each cluster. While our interest was mainly directed to the Parkinson's classification problem, the methods seem relatively generic and should be applicable to many data sets. (As an example, we also applied this directly to a well-known baseline data set - wine classification and obtained state of the art results). On the Parkinson classification task, these methods obtained, on a 7 degree classification scale, results which are comparable to the best accuracy on simple two class classification.

A novel fMRI classification method designed for rapid event related fMRI experiments is described and applied to the classification of loud reading of isolated words in Hebrew. Three comparisons of different grammatical complexity were performed: (i) words versus asterisks (ii) “with diacritics versus without diacritics” and (iii) “with root versus no root”. We discuss the most difficult task and, for comparison, the easiest one. Earlier work using more standard classification techniques (machine learning and statistical) succeeded fully only in the simplest of these tasks (i), but produced only partial results on (ii) and failed completely, even on the training set on the deepest task (iii). The method performs a “best match” between pre-processed data and computing a full library of artificially generated examples. The method involves a deconvolution of the rapid events on the data and performing a convolution on the generated data. The best-match is performed over all “words” constructed by convolving the response functions of each value of each event performed in a “windowed” sequence. This is accomplished separately for all voxels and then a voting procedure defines the outcome. Using the same feature selection (ANOVA) as in the earlier methods, (i) there is a dramatic increase in the accuracy rate for the third (most difficult task) on the intra-run level (88%) as well as the first task (ii) Unlike the earlier methods training and testing over all runs (within subject) achieves a significant level of classification (64% accuracy) for the training set. This shows the information for this “deeper” cognitive task can in fact be extracted from the fMRI information.

BACKGROUND: Classification using microarray datasets is usually based on a small number of samples for which tens of thousands of gene expression measurements have been obtained. The selection of the genes most significant to the classification problem is a challenging issue in high dimension data analysis and interpretation. A previous study with SVM-RCE (Recursive Cluster Elimination), suggested that classification based on groups of correlated genes sometimes exhibits better performance than classification using single genes. Large databases of gene interaction networks provide an important resource for the analysis of genetic phenomena and for classification studies using interacting genes.We now demonstrate that an algorithm which integrates network information with recursive feature elimination based on SVM exhibits good performance and improves the biological interpretability of the results. We refer to the method as SVM with Recursive Network Elimination (SVM-RNE) RESULTS: Initially, one thousand genes selected by t-test from a training set are filtered so that only genes that map to a gene network database remain. The Gene Expression Network Analysis Tool (GXNA) is applied to the remaining genes to form n clusters of genes that are highly connected in the network. Linear SVM is used to classify the samples using these clusters, and a weight is assigned to each cluster based on its importance to the classification. The least informative clusters are removed while retaining the remainder for the next classification step. This process is repeated until an optimal classification is obtained. CONCLUSION: More than 90% accuracy can be obtained in classification of selected microarray datasets by integrating the interaction network information with the gene expression information from the microarrays.The Matlab version of SVM-RNE can be downloaded from http://web.macam.ac.il/~myousef.

Collaborative Filtering (CF) is currently one of the most popular and most widely used personalization techniques. It generates personalized predictions based on the assumption that users with similar tastes prefer similar items. One of the major drawbacks of the CF from the computational point of view is its limited scalability since the computational effort required by the CF grows linearly both with the number of available users and items. This work proposes a novel efficient variant of the CF employed over a multidimensional content-addressable space. The proposed approach heuristically decreases the computational effort required by the CF algorithm by limiting the search process only to potentially similar users. Experimental results demonstrate that the proposed heuristic approach is capable of generating predictions with high levels of accuracy, while significantly improving the performance in comparison with the traditional implementations of the CF.

In this paper, we show how a simple feed-forward neural network can be trained to filter documents when only positive information is available, and that this method seems to be superior to more standard methods, such as tf-idf retrieval based on an “average vector”. A novel experimental finding that retrieval is enhanced substantially in this context by carrying out a certain kind of uniform transformation (“Hadamard”) of the information prior to the training of the network.

It is shown how a single homogeneous SDM memory can be organized to link between low level information and high level correlations. To illustrate this, we report on experiments run in a unified memory retrieval system, that combined pattern recognition of individual English characters followed by the assignment of ‘meaning’ to a string by giving it a Hebrew translation. Symmetry allows the reverse action on the same memory (i.e. Hebrew character identification followed by translation of a string to English).

In many applications, the use of Bayesian probability theory is problematical. Information needed to feasibility calculate is unavailable. There are different methodologies for dealing with this problem, e.g., maximal entropy and Dempster-Shafer Theory. If one can make independence assumptions, many of the problems disappear, and in fact, this is often the method of choice even when it is obviously incorrect. The notion of independence is a 0–1 concept, which implies that human guesses about its validity will not lead to robust systems. In this paper, we propose a fuzzy formulation of this concept. It should lend itself to probabilistic updating formulas by allowing heuristic estimation of the “degree of independence.” We show how this can be applied to compute a new notion of conditional probability (we call this “extended conditional probability”). Given information, one typically has the choice of full conditioning (standard dependence) or ignoring the information (standard independence). We list some desiderata for the extension of this to allowing degree of conditioning. We then show how our formulation of degree of independence leads to a formula fulfilling these desiderata. After describing this formula, we show how this compares with other possible formulations of parameterized independence. In particular, we compare it to a linear interpolant, a higher power of a linear interpolant, and to a notion originally presented by Hummel and Manevitz [Tenth Int. Joint Conf. on Artificial Intelligence, 1987]. Interestingly, it turns out that a transformation of the Hummel-Manevitz method and our “fuzzy” method are close approximations of each other. Two examples illustrate how fuzzy independence and extended conditional probability might be applied. The first shows how linguistic probabilities result from treating fuzzy independence as a linguistic variable. The second is an industrial example of troubleshooting on the shop floor.

A method for constructively approximating functions in the uniform (i.e., maximal error) norm by successive changes in the weights and number of neurons in a neural network is developed. This is a realization of the approximation results of Cybenko, Hecht-Nielsen, Hornik, Stinchcombe, White, Gallant, Funahashi, Leshno et al., and others. The constructive approximation in the uniform norm is more appropriate for a number of examples, such as robotic arm motion, and stands in contrast with more standard methods, such as back-propagation, which approximate only in the average error norm. Copyright 1996 Elsevier Science Ltd

Reasoning with uncertainty is a field with many different approaches and viewpoints, with important applications to sensor design and autonomous system development. It is important to have calculi for propagating measures of "probability" or "likelihood" even in cases of subjective information, and it is just as important to be able to propagate the "certitude" of this information. By choosing the semantics properly, this information can be handled by keeping track of certain statistics on a different probability space, (which we call the opinion space). The semantics assume that the "likelihood" or "probability numbers" are in fact averages over many (perhaps subjective) opinions and that uncertainty is represented by the spread in these opinions, which can be technically maintained by a covariance matrix. Different calculi result from different design choices consistent with this choice of semantics. It also turns out that certain mechanisms that are frequently considered "non-Bayesian", result from specific choices for representing the statistics and dependency assumptions.

Determining the truth value of self-referential sentences is an interesting and often tricky problem. The Gödel sentence, asserting its own unprovability in P (Peano arithmetic), is clearly true in N(the standard model of P), and Löb showed that a sentence asserting its own provability in P is also true in N (see Smorynski [Sm, 4.1.1]). The problem is more difficult, and still unsolved, for sentences of the kind constructed by Kreisel [K1], which assert their own falsity in some model N* of P whose complete diagram is arithmetically defined. Such a sentence χ has the property that N ⊨ iff N* ⊭ χ (note that ¬χ has the same property). We show in §1 that the truth value in N of such a sentence χ, after a certain normalization that breaks the symmetry between it and its negation, is determined by the parity of a natural number, called the rank of N, for the particular construction of N* used. The rank is the number of times the construction can be iterated starting from N and is finite for all the usual constructions. We also show that modifications of, e.g., Henkin's construction (in his completeness proof of predicate calculus) allow arbitrary finite values for the rank of N. Thus, on the one hand the truth value of χ in N, for a given “nice” construction of N*, is independent of the particular (normalized) choice of χ, and we shall see that χ is unique up to (provable) equivalence in P. On the other hand, the truth value in question is sensitive to minor changes in the definition of N* and its determination seems to be largely a combinatorial problem.