Navegando por Autor "Assis, Alex Damiany"

Agora exibindo 1 - 2 de 2

Fast placement and routing by extending coarse-grained reconfigurable arrays with Omega Networks.
(2011) Ferreira, Ricardo dos Santos; Cardoso, João Manuel Paiva; Assis, Alex Damiany; Vendramini, Júlio; Teixeira, Tiago
Reconfigurable computing architectures are commonly used for accelerating applications and/or for achieving energy savings. However, most reconfigurable computing architectures suffer from computationally demanding placement and routing (P&R) steps. This problem may disable their use in systems requiring dynamic compilation (e.g., to guarantee application portability in embedded systems). Bearing in mind the simplification of P&R steps, this paper presents and analyzes a coarse-grained reconfigurable array (CGRA) extended with global multistage interconnect networks, specifically Omega Networks. We show that integrating one or two Omega Networks in a CGRA permits to simplify the P&R stage resulting in both low hardware resource overhead and low performance degradation (18% for an 8 _ 8 array). We compare the proposed CGRA, which integrates one or two Omega Networks, with a CGRA based on a grid of processing elements with reach neighbor interconnections and with a torus topology. The execution time needed to perform the P&R stage for the two array architectures shows that the array using two Omega Networks needs a far simpler and faster P&R. The P&R stage in our approach completed on average in about 16_ less time for the 17 benchmarks used. Similar fast approaches needed CGRAs with more complex interconnect resources in order to allow most of the benchmarks used to be successfully placed and routed.
Neural networks regularization with graph-based local resampling.
(2021) Assis, Alex Damiany; Torres, Luiz Carlos Bambirra; Araújo, Lourenço Ribeiro Grossi; Hanriot, Vítor Mourão; Braga, Antônio de Pádua
This paper presents the concept of Graph-based Local Resampling of perceptron-like neural networks with random projections (RN-ELM) which aims at regularization of the yielded model. The addition of synthetic noise to the learning set finds some similarity with data augmentation approaches that are currently adopted in many deep learning strategies. With the graph-based approach, however, it is possible to direct resample in the margin region instead of exhaustively cover the whole input space. The goal is to train neural networks with added noise in the margin region, located by structural information extracted from a planar graph. The so-called structural vectors, which are the training set vertices near the class boundary, are obtained from the structural information using Gabriel Graph. Synthetic samples are added to the learning set around the geometric vectors, improving generalization performance. A mathematical formulation that shows that the addition of synthetic samples has the same effect as the Tikhonov regularization is presented. Friedman and pos-hoc Nemenyi tests indicate that outcomes from the proposed method are statistically equivalent to the ones obtained by objective-function regularization, implying that both methods yield smoother solutions, reducing the effects of overfitting.