Significant Reduction Of Computational Load Research Articles

A machine-learning-based event-triggered model predictive control for building energy management

Model predictive control (MPC) for building energy management has exhibited a huge potential for largely cutting down energy use and improving human comfort. However, the high demand for computational power for solving the optimization is challenging the widespread deployment of MPC in real buildings. Typical MPC employs a time-triggered mechanism (TTM) that runs optimization recurrently at each time step without considering the necessity, which could cause excessive computation resource usage. This paper proposes an event-triggered mechanism (ETM) that only runs optimization when triggering events occur. Contrary to conventional ETMs that only considers the current information (e.g., room conditions), the ETM proposed is based on a cost function that covers the past, current, and future information. Based on the proposed ETM, a machine-learning-based event-triggered model predictive control (ETMPC) system that optimizes both building energy efficiency and thermal comfort is developed. The developed ETMPC system is then employed for air-conditioning control for performance evaluation through simulations. The control performance of the proposed ETMPC is compared to an MPC employing TTM and a common thermostat. Compared to the MPC employing TTM, the proposed ETMPC reduced the computational load in terms of the number of optimization runs by 77.6%–88.2%, meanwhile, achieving similar energy savings (more than 9.3% energy saving over the thermostat) as the MPC employing TTM does (9% energy saving over the thermostat). With the significant reduction of computational load, the ETMPC achieves similar thermal comfort performance as the MPC employing TTM does, and it is significantly better over the thermostat.

Model predictive consensus in networked autonomous systems using discrete Laguerre functions and event triggering approach

This paper presents the consensus problem of general linear discrete time networked autonomous agents, using distributed model predictive control (DMPC) and event triggering algorithm. The computational complexity of conventional DMPC is high due to the large values of prediction horizon and control horizon required for real time systems. The increased computational load becomes an obstruction in the practical implementation of DMPC. To solve the issue, a novel design methodology of parameterizing the control input using discrete Laguerre functions is adopted in this paper. The predictive control protocol is applied to a group of networked autonomous agents operating on event triggered consensus algorithm. The proposed event triggered algorithm offers the advantages of lesser communication between the agents along with reduction in controller updates. The closed loop stability of the networked system is established using Lyapunov criterion, based on the terminal state’s equality constraint. Comparison of the proposed algorithm with existing works through numerical examples shows significant reduction in computational load, design parameters and data transmission, while achieving comparable performance. To validate the theoretical results, simulation is carried out on a group of autonomous vehicles and the usefulness of the proposed algorithm on system performance is illustrated.

Inference Skipping for More Efficient Real-Time Speech Enhancement With Parallel RNNs

Deep neural network (DNN) based speech enhancement models have attracted extensive attention due to their promising performance. However, it is difficult to deploy a powerful DNN in real-time applications because of its high computational cost. Typical compression methods such as pruning and quantization do not make good use of the data characteristics. In this paper, we introduce the Skip-RNN strategy into speech enhancement models with parallel RNNs. The states of the RNNs update intermittently without interrupting the update of the output mask, which leads to significant reduction of computational load without evident audio artifacts. To better leverage the difference between the voice and the noise, we further regularize the skipping strategy with voice activity detection (VAD) guidance, saving more computational load. Experiments on a high-performance speech enhancement model, dual-path convolutional recurrent network (DPCRN), show the superiority of our strategy over strategies like network pruning or directly training a smaller model. We also validate the generalization of the proposed strategy on two other competitive speech enhancement models.

An optimization‐based approach for structural design of self‐assembled DNA tiles

DNA tiles are self‐assembled nanostructures, which offer exciting opportunities for synthesis of novel materials. A challenge for structural design of DNA tiles is to identify optimal locations for so‐called crossovers, which are bridges between DNA double helices formed by pairs of single‐stranded DNA. An optimization‐based approach is presented to identify optimal locations for such crossovers. Minimization of a potential‐energy model for a given structural design demonstrates the importance of local minima. Both deterministic global optimization of a reduced model and multistart optimization of the full model are applied successfully to identify the global minimum. MINLP optimization using a branch‐and‐bound algorithm (GAMS/SBB) identifies an optimal structural design of a DNA tile successfully with significant reduction in computational load compared to exhaustive enumeration, which demonstrates the potential of the proposed method to reduce trial‐and‐error efforts for structural design of DNA tiles. © 2016 American Institute of Chemical Engineers AIChE J, 63: 1804–1817, 2017

Comparison of feature selection methods using ANNs in MCP-wind speed methods. A case study

Recent studies in the field of renewable energies, and specifically in wind resource prediction, have shown growing interest in proposals for Measure–Correlate–Predict (MCP) methods which simultaneously use data recorded at various reference weather stations. In this context, the use of a high number of reference stations may result in overspecification with its associated negative effects. These include, amongst others, an increase in the estimation error and/or overfitting which could be detrimental to the generalisation capacity of the model when handling new data (prediction).This paper analyses the benefits of feature selection for use with Artificial Neural Network (ANN) techniques with a multilayer perceptron (MLP) structure when the ANNs are used as MCP methods to predict mean hourly wind speeds at a target site. The features considered in this study were the mean hourly wind speeds and directions recorded in 2003 and 2004 at five weather stations in the Canary Archipelago (Spain).The two feature selection techniques considered in the analysis were the Correlation Feature Selection (CFS), which is a correlation-based filter approach (FA), and an MLP-based wrapper approach (WA). The metrics used to compare the results were the mean absolute error (MAE), the mean absolute percentage error (MAPE) and the index of agreement (IoA).Evaluation of the mean errors obtained in the 10-fold cross-validation tests for the year used to represent the short-term wind data period resulted in several conclusions. These included, notably, that the WA gave lower mean errors than the FA in 100% of the cases analysed independently of the metric employed. However, the FA resulted in a significant reduction in computational load and considerable enhancement of model interpretability. When very good correlation coefficients were obtained between the target and reference stations, no significant statistical difference was observed at 5% level between the three models (FA, WA and the models constructed with all the variables) in most of the cases analysed.

A Single Symbol Maximum-Likehood Decoder for Four Transimit Antenna QOSTBC

This letter proposes a very low-complexity maximum likelihood (ML) detection algorithm for the quasi-orthogonal space-time block code (QOSTBC) with four transmits antennas. The low-complexity single-complex-symbol detection is based on QR decomposition and the expansion of equalization via a diagonal matrix transformation. The proposed strategy enables the QOSTBC to achieve ML performance with significant reduction in computational load for any high-level modulation scheme.

Dynamic capacity control in cargo revenue management – A new heuristic for solving the single-leg problem efficiently

While seat inventory control is a standard technique in passenger revenue management (RM) to steer availability and thus prices of passenger seats, far less attention has been paid to dynamically managing cargo bookings. Two technical challenges of cargo RM that drive this contrast are the multi-dimensional nature of cargo shipments and request sizes featuring multiple capacity units. Because of these characteristics, the exact decision problem is computationally intractable. This article proposes a heuristic for solving the air cargo capacity control problem efficiently while obtaining near-optimal solutions. A significant reduction of computational load is achieved through exploiting the structure of monotone switching curves.

Low-Complexity Maximum-Likelihood Decoder for Four-Transmit- Antenna Quasi-Orthogonal Space–Time Block Code

This letter proposes a very low-complexity maximum-likelihood (ML) detection algorithm based on QR decomposition for the quasi-orthogonal space-time block code (QSTBC) with four transmit antennas, called the LC-ML decoder. The proposed algorithm enables the QSTBC to achieve ML performance with significant reduction in computational load for any high-level modulation scheme.

Very-Low-Complexity Maximum-Likelihood Decoders for Four-Transmit-Antenna Quasi-Orthogonal Space-Time Code

This letter proposes two very-low-complexity maximum-likelihood (ML) detection algorithms based on QR decomposition for the quasi-orthogonal space-time code (QSTBC) with four transmit antennas [3]-[5], called VLCMLDec1 and VLCMLDec2 decoders. The first decoder, VLCMLDecl, can be used to detect transmitted symbols being extracted from finite-size constellations such as phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The second decoder, VL-CMLDec2, is an enhanced version of the VLCMLDecl, developed mainly for QAM constellations. Simulation results show that both of the proposed decoders enable the QSTBC to achieve ML performance with significant reduction in computational load.

Fast<tex>$k$</tex>-NN Classification Using the Cluster-Space Approach

A fast k-nearest neighbor algorithm is presented which combines k-NN with a cluster-space data representation. Implementation of the algorithm is easier, and classification time can be significantly reduced. Computer-generated data show the modified k-NN retains the advantage of nonparametric analysis but with significant reduction in computational load. Results from tests carried out with Hyperion data demonstrate that the simplification has little effect on classification performance, and yet efficiency is greatly improved.