IEEE Transactions on Neural Networks

By G. Costantini, Renzo Perfetti, Massimiliano Todisco. 2008

A new neural network for convex quadratic optimization is presented in this brief. The proposed network can handle both equality and inequality constraints, as well as bound constraints on the optimization variables. It is based on the Lagrangian approach, but exploits a partial dual method in order to keep the number of variables at minimum. The dynamic evolution is globally convergent and the steady-state solutions satisfy the necessary and sufficient conditions of optimality. The circuit implementation is simpler with respect to existing solutions for the same class of problems. The validity of the proposed approach is verified through some simulation examples.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Jing Ma
2. Wei Gao
3. Junbo Zhang
4. Kam-Fai Wong
5. Xiuwen Yi
6. Jindong Wang
7. Yu Zheng
8. Yuxuan Liang

Citations in the corpus

• Right for the Right Reasons: Training Differentiable Models by Constraining their Explanations (2017, background): “At the other extreme, whenA is a matrix of all 1s, it encourages the model to have small gradients with respect to its inputs; this can improve generalization on its own [Drucker and Le Cun, 1992].”
• A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction (2017, background): “Traditional RNNs, however, suffer from the problem of vanishing gradients [Bengio et al., 1994] and thus have difficulty capturing long-term dependencies.”
• Detecting Rumors from Microblogs with Recurrent Neural Networks (2016, background): “In practice, because of the vanishing or exploding gradients [Bengio et al., 1994], the basic RNN cannot learn long-distance temporal dependencies with gradient-based optimization.”
• Learning to Solve NP-Complete Problems - A Graph Neural Network for the Decision TSP (2018, methodology): “A promising meta-architecture to engineer models that learn on symbolic domains is to instantiate neural modules and assemble them in various configurations, each manifesting a graph representation of a given instance of the problem at hand (Scarselli et al. 2009).”
• GeoMAN: Multi-level Attention Networks for Geo-sensory Time Series Prediction (2018, methodology): “Moreover, we use stacked LSTMs (the number of layers is denoted as q ) as the unit of encoder and decoder to enhance our performance.”
• Towards Explainable Text Classification by Jointly Learning Lexicon and Modifier Terms (2017, methodology): “…algorithms [Fernández-Delgado et al., 2014] have helped improve document navigation tasks by categorizing web search results [Chen and Dumais, 2000], analyzed corpora to identify anonymous authors [Diederich et al., 2003], and are used to identify spam e-mails [Drucker et al., 1999] at large scale.”
• Analysis of the emission of very small dust particles from Spitzer spectro-imagery data using blind signal separation methods (2007, methodology): “The original Basic Calibrated Data (BCD) files are from the S12 pipeline of theSpitzer Science Center.”
• FedHealth: A Federated Transfer Learning Framework for Wearable Healthcare (2019, background): “, 2012], or distribution alignment to reduce the marginal or conditional distribution divergence between domains [Wang et al., 2019b; Wang et al., 2018a; Wang et al., 2017a; Pan et al., 2011; ?].”
• Graph Classification via Deep Learning with Virtual Nodes (2017, methodology): “Yet recurrent nets are also employed along the random walk from a node [Scarselli et al. , 2009].”