Browsing by Subject "Neural networks"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Publication
    Open Access
    Análisis mediante inteligencia artificial de las emociones del alumnado autista en la interacción social con el robot NAO
    (Universidad de Murcia, Servicio de Publicaciones, 2024-05-30) Lorenzo Lledó, Gonzalo; Lorenzo Lledó, Alejandro; Rodríguez Quevedo, Ángel
    Actualmente, la tecnología es la herramienta más utilizada en el desarrollo de las actividades de la vida diaria. Cada vez es mayor, el número de campos de conocimiento que se benefician de su versatilidad y la aplicación en el desarrollo de sus actividades. En el entorno educativo, permite generar actividades adaptadas a las necesidades del alumnado. En los últimos años, la robótica y la inteligencia artificial son las que mayor difusión están teniendo. Las características de estas herramientas favorecen su aplicación con el alumnado con Trastorno del Espectro Autista. Por tanto, el objetivo de la investigación es la aplicación de la robótica para favorecer la comunicación e interacción social en el alumnado con autismo analizando las emociones que manifiestan a lo largo de las distintas actividades. Para ello, se implementó un estudio piloto con el robot NAO y cuatro niños autistas que desarrollaron actividades de imitación, juego e interacción social. Durante su realización se utilizó un sistema automático basado en redes neuronales convolucionales para detectar los estados de ánimo en el proceso de interacción. Los resultados muestran que tristeza, felicidad y enfado son las emociones que tiene una mayor probabilidad de producirse en los participantes. Por tanto, se concluye que el robot y el sistema de inteligencia artificial son un elemento fundamental para ayudar a expresar sus emociones en las interacciones sociales.
  • Thumbnail Image
    Publication
    Restricted
    Compiler-Assisted Instruction Fusion
    (IEEE, 2026) Reddy, Ravikiran Ravindranath; Singh, Sawan; Perais, Arthur; Ros Bardisa, Alberto; Jimborean, Alexandra; Ingeniería y Tecnología de Computadores
    Hardware instruction fusion combines multiple architectural instructions into a single operation, improving performance by freeing up resources. While fusion typically involves consecutive instructions, there are proposals to fuse nonconsecutive instructions to maximize potential. However, such approaches require complex and costly hardware to predict and either validate fusion or unfuse, which significantly increases the cost of fusion. In this work, we propose a compiler technique, CAIF - Compiler Assisted Instruction Fusion, for fusionaware instruction scheduling. CAIF identifies fusible but nonconsecutive memory operations and reorders eligible pairs of instructions such that they appear consecutively in the instruction stream. Our experiments demonstrate that for neural network workloads, a hardware that only fuses consecutive instructions obtains 1.2% average performance improvements over a no-fusion baseline when applications are compiled with a standard compiler and 19.6% when compiled with CAIF. In addition, when nonconsecutive hardware fusion (Helios) is enabled, CAIF boosts performance from 6.6% to 20.3%. Moreover, CAIF can effectively handle the statically challenging general-purpose application and boost performance on SPEC CPU 2017 from 2.4% to 6.4%, and from 14.4% to 17.7%, respectively, on the hardware configurations mentioned above.
  • Thumbnail Image
    Publication
    Open Access
    Convolutional neural networks for estimating the ripening state of fuji apples using visible and near-infrared spectroscopy
    (Springer , 2022-07-18) Benmouna, Brahim; García Mateos, Ginés; Sabzi, Sajad; Fernández Beltrán, Rubén; Parras Burgos, Dolores; Molina Martínez, José Miguel; Informática y Sistemas; Facultades de la UMU::Facultad de Informática
    The quality of fresh apple fruits is a major concern for consumers and manufacturers. Classification of these fruits according to their ripening stage is one of the most decisive factors in determining their quality. In this regard, the aim of this work is to develop a new method for non-destructive classification of the ripening state of Fuji apples using hyperspectral information in the visible and near-infrared (Vis/NIR) regions. Spectra of 172 apple samples in the range from 450 to 1000 nm were studied, which were selected from four different ripening stages. A convolutional neural network (CNN) model was proposed to perform the classification of the samples. The proposed method was compared with three alternative methods based on artificial neural networks (ANN), support vector machines (SVM), and k-nearest neighbors (KNN). The results revealed that the CNN method outperformed the alternative methods, achieving a correct classification rate (CCR) of 96.5%, compared with an average of 89.5%, 95.93%, and 91.68% for ANN, SVM, and KNN, respectively. These results will help in the development of a new device for fast and accurate estimation of the quality of apples.
  • Thumbnail Image
    Publication
    Open Access
    Enabling Automatic Compiler-Driven Vectorization of Transformers
    (IEEE, 2026-02-23) Alladi, Shreya; Ros Bardisa, Alberto; Jimborean, Alexandra; Ingeniería y Tecnología de Computadores
    Compiling neural networks and Transformers for edge devices faces significant challenges due to resource constraints and the reliance on manually optimized operations for performance among others. These limitations hinder the scalability and portability of neural networks on resource-constrained platforms, such as edge devices utilizing the RISC-V ecosystem. Addressing these issues, this paper introduces innovative techniques to overcome the inefficiencies of current compilation methods and reduce dependence on manual optimizations. This work proposes a novel compilation flow, ONNXMLIR- LLVM (OML), which leverages MLIR and LLVM IR to enable automatic optimizations and generate stand-alone RISC-V binaries. Through comprehensive analysis, we identify key barriers preventing the auto-vectorizer from handling vectorization-friendly operators, particularly reduction operations and vectorization-unfriendly data layouts. We address these through a versatile MLIR reduction detection pass and a compiletime transpose pass, respectively. Our automatic transformations (OML-vect) unlock the capabilities of the MLIR affine super-vectorizer, reducing reliance on manual vectorization. Evaluations on both x86 and RISC-V across eight neural networks and Transformer models demonstrate that automatic vectorization via OML-vect achieves, on average, 5% and 59% on x86 and RISC-V, respectively, compared to baseline (manually vectorized libraries), offering an efficient and portable solution for edge device deployments.
  • Thumbnail Image
    Publication
    Open Access
    Histological heterogeneity of human glioblastomas investigated with an unsupervised neural network (SOM)
    (Murcia : F. Hernández, 2005) Iglesias-Rozas, J.R.; Hopf, N.
    The histological variability of Glioblastomas (GB) precludes the modern assimilation of theses tumors into a single histological tumor group. As an alternative to statistical histological evaluation, we investigated 1489 human GB in order to discover whether they could be correctly classified using Self- Organizing Maps (SOM). In all tumors 50 histological features, as well as the age and sex of the patients, were examined. Four clusters of GB with a significance of 52 (maximal significance 60) were found. Cluster C1 contained 37.47% of all GB and 41.09% of all polymorphic glioblastomas (PG). Cluster C2 included 35.06% of all GB and 44.96% of all giant cell glioblastomas (GCG). Cluster C3 contained 16.45% of all GB with a significant component of astroblasts, glioblasts and oligodendroglia. Cluster C4 included 11.01% of all GB, 87.80% of the gliosarcomas (GS) and 36.72% of all GCG. Placing a series of component windows with their maps side by side allows the immediate recognition of the dependencies on variables and the determination of variables necessary to build the specific clusters. The SOM allow a realistic histological classification, comparable to the actual classification by the WHO. In addition, we found new, small subclusters of human GB which may have a clinical significance. With SOM one can learn to discriminate, discard and delete data, select histological and clinical or genetic variables that are meaningful, and consequently influence the result of patient management.