NETWORKS GOVERNED BY ŁUKASIEWICZ LOGICS Leandro, C a ; Pita, H b ; Monteiro, L c

a_{Área Científica da Matemática, Instituto Superior de Engenharia de Lisboa, Instituto}

Politécnico de Lisboa, Portugal

b_{ISEL, Inst Super Engn Lisboa, Departamento de Engenharia Electrónica Telecomunicações e}

de Computadores, P-1959-007, Lisbon, Portugal

c_{Departamento de Informática, Faculdade de Ciências e Tecnologia, Universidade Nova de}

Lisboa, Portugal

Fonte: Computational Intelligence Book Series: Studies in Computational Intelligence, Volume 343, Pages 45-58, 2011

Conferência: 1st International Joint Conference on Computational Intelligence, October 05- 07, 2009, Funchal, Portugal

ISSN: 1860-949X

ISBN: 978-3-642-20205-6 Editor: Springer-Verlag Berlin

Tipo de Documento: Proceeding Paper Área Científica: Computer Science

Resumo: This work describes a methodology to extract symbolic rules from trained neural networks. In our approach, patterns on the network are codified using formulas on a Łukasiewicz logic. For this we take advantage of the fact that every connective in this multi- valued logic can be evaluated by a neuron in an artificial network having, by activation function the identity truncated to zero and one. This fact simplifies symbolic rule extraction and allows the easy injection of formulas into a network architecture. We trained this type of neural network using a back-propagation algorithm based on Levenderg-Marquardt algorithm, where in each learning iteration, we restricted the knowledge dissemination in the network structure. This makes the descriptive power of produced neural networks similar to the descriptive power of Łukasiewicz logic language, minimizing the information loss on the translation between connectionist and symbolic structures. To avoid redundance on the generated network, the method simplifies them in a pruning phase, using the “Optimal Brain Surgeon” algorithm. We tested this method on the task of finding the formula used on the generation of a given truth table. For real data tests, we selected the Mushrooms data set, available on the UCI Machine Learning Repository.

THIN-FILM PHOTODIODE WITH AN a-Si:H/nc-Si:H ABSORPTION BILAYER Vygranenko, Y.a,b,c_{; Vieira, M.}a,b_{; Sazonov, A.}a,c

a_{ISEL, Inst Super Engn Lisboa, Electronics Telecommunications and Computer Engineering,}

1959-007, Lisbon, Portugal

b_{CTS-UNINOVA, 2829-516 Caparica, Portugal}

Fonte: MRS Online Proceedings Library, Volume 1321, Pages 11-1321-a20-06, 2011 DOI: 10.1557/opl.2011.952

Editor: Materials Research Society Tipo de Documento: Article

Resumo: We report on the fabrication and characterization of n+-n-i-δi-p thin-film photodiodes with an active region comprising a hydrogenated nanocrystalline silicon (nc- Si:H) n-layer and a hydrogenated amorphous silicon (a-Si:H) i-layer. The combination of wide- and narrow-gap absorption layers enables the spectral response extending from the near-ultraviolet (NUV) to the near-infrared (NIR) region. Moreover, in the low-bias range, when only the i-layer is depleted, the leakage current is significantly lower than that in the conventional nc-Si:H n+-n-p+ photodiode deposited under the same deposition conditions. Device with the 900nm/400nm thick n-i-layers exhibits a reverse dark current density of 3 nA/cm2 _{at −1V. In the high-bias range, when the depletion region expands within the n-layer,}

the magnitude of the leakage current depends on electronic properties of nc-Si:H. The density of shallow and deep states, and diffusion length of holes in the n-layer have been estimated from the capacitance-voltage characteristics and from the bias dependence of the long- wavelength response, respectively. To improve the quantum efficiency in the NIR-region, we have also implemented a Cr / ZnO:Al back reflector. The observed long-wavelength spectral response is about twice as high as that for a reference photodiode without ZnO:Al layer. Results demonstrate the feasibility of the photodiode for low-level light detection in the NUV-to-NIR spectral range.

THREE CURRENT ISSUES IN MUSIC AUTOTAGGING Marques, G.a_{; Domingues, M.}b_{, Langlois, T.}c_{; Gouyon, F.}b

a_{ISEL, Inst Super Engn Lisboa, DEETC P-1959-007, Lisbon, Portugal} b_{INESC Porto, Portugal}

c_{DI-FCUL Lisboa, Portugal}

Fonte: Proceedings of the 12th International Society for Music Information Retrieval, Pages 795-800, 2011

Conference: 12th International Society for Music Information Retrieval (ISMIR 2011), Miami, USA, October 10, 2011

Editor: Anssi Klapuri and Colby Leiber Tipo de Documento: Proceeding Paper

Resumo: The purpose of this paper is to address several aspects of music autotagging. We start by presenting autotagging experiments conducted with two different systems and show performances on a par with a method representative of the state-of-the-art. Beyond that, we illustrate via systematic experiments the importance of a number of issues relevant to autotagging, yet seldom reported in the literature. First, we show that the evaluation of autotagging techniques is fragile in the sense that small alterations to the set of tags to be learned, or in the set of music pieces may lead to dramatically different results. Hence we stress a set of methodological recommendations regarding data and evaluation metrics. Second, we conduct experiments on the generality of autotagging models, showing that a number of different methods at a similar performance level to the state-of-the-art fail to learn tag models able to generalize to datasets from different origins. Third we show that current performance level of a direct mapping between audio features and tags still appears

insufficient to enable the possibility of exploiting natural tag correlations as a second stage to improve performance.

THREE TRANSDUCERS EMBEDDED INTO ONE SINGLE SiC

PHOTODETECTOR: LSP DIRECT IMAGE SENSOR, OPTICAL AMPLIFIER AND