This paper studies the performative prediction problem where a learner aims to minimize the expected loss with a decision-dependent data distribution. Such setting is motivated when outcomes can be affected by the pre...
详细信息
Actions taken by building occupants and facility managers can have significant impacts on building energy performance. Despite the growing interesting in understanding human drivers of energy consumption, literature o...
详细信息
ISBN:
(纸本)9781509044849
Actions taken by building occupants and facility managers can have significant impacts on building energy performance. Despite the growing interesting in understanding human drivers of energy consumption, literature on the topic remains limited and is mostly focused on studying individual occupancy actions (e.g., changing thermostat set point temperatures). Consequently, the impact of uncertainty in human actions on overall building performance remains unclear. This paper proposes a novel method to quantify the impact of potential uncertainty in various operation actions on building performance, using a combination of Monte Carlo and Fractional Factorial analyses. The framework is illustrated in a case study on educational buildings, where deviations from base case energy intensity levels exceed 50 kWh/m 2 /year in some cases. The main contributors to this variation are the thermostat temperature set point settings, followed by the consumption patterns of equipment and lighting systems by occupants during unoccupied periods.
A data-mining approach is proposed in this study for estimating the interval cycle time of each job in a semiconductor manufacturing system, which was seldom investigated in the past studies but is a critical task to ...
详细信息
The challenges in applying contrastive learning to speaker verification (SV) are that the softmax-based contrastive loss lacks discriminative power and that the hard negative pairs can easily influence learning. To ov...
详细信息
The challenges in applying contrastive learning to speaker verification (SV) are that the softmax-based contrastive loss lacks discriminative power and that the hard negative pairs can easily influence learning. To overcome the first challenge, we propose a contrastive learning SV framework incorporating an additive angular margin into the supervised contrastive loss in which the margin improves the speaker representation’s discrimination ability. For the second challenge, we introduce a class-aware attention mechanism through which hard negative samples contribute less significantly to the supervised contrastive loss. We also employed gradient-based multi-objective optimization to balance the classification and contrastive loss. Experimental results on CN-Celeb and Voxceleb1 show that this new learning objective can cause the encoder to find an embedding space that exhibits great speaker discrimination across languages.
A priority queueing model with many types of requests and restricted processor sharing is considered. A novel discipline of requests admission and service is proposed. This discipline assumes restriction of the bandwi...
详细信息
This study analyses the critical success factors (CSFs) of construction supply chain management (CSCM) using a mix-method approach to identify and prioritise the CSFs for CSCM performance. An extensive literature revi...
详细信息
Recurrent neural network language models (RNNLM) have become an increasingly popular choice for state-of-the-art speech recognition systems. RNNLMs are normally trained by minimizing the cross entropy (CE) using the s...
详细信息
ISBN:
(纸本)9781538646595
Recurrent neural network language models (RNNLM) have become an increasingly popular choice for state-of-the-art speech recognition systems. RNNLMs are normally trained by minimizing the cross entropy (CE) using the stochastic gradient descent (SGD) algorithm. The SGD method only uses first-order derivatives and no higher order gradient information is used to consider the correlation between model parameters. It is unable to fully capture the curvature of the error cost function. This can lead to slow convergence in model training. In this paper, a limited-memory Broyden Fletcher Goldfarb Shannon (L-BFGS) based second order optimization technique is proposed for RNNLMs. This method efficiently approximates the matrix-vector product between the inverse Hessian and gradient vector via a recursion over past gradients with a compact memory requirement. Consistent perplexity and error rate reductions are obtained over the SGD method on two speech recognition tasks: Switchboard English and Babel Cantonese. A faster convergence and speed up in RNNLM training time was also obtained. Index Terms: recurrent neural network, language model, second order optimization, limited-memory BFGS, speech recognition.
The hybrid self-organization map (SOM) and radial basis function (RBF) network approach is proposed in this study for forecasting the interval cycle time of each job in a wafer fabrication factory (wafer fab), which i...
详细信息
With a significant number of states in the U.S. trading electricity in restructured markets, a significant proportion of capacity expansion in the future will have to take place in market based environments. However, ...
详细信息
With a significant number of states in the U.S. trading electricity in restructured markets, a significant proportion of capacity expansion in the future will have to take place in market based environments. However, since a majority of the literature on capacity expansion is focused on regulated market structures, there is a critical need for comprehensive multiperiod, multi-player, capacity expansion models. In this research, we present a 3-tier game theoretic model to obtain multi-period, multi-player equilibrium capacity expansion plans, while considering several relevant market features. We will develop a reinforcement learning based approach to obtain the model solution. The model will be tested on sample power networks and benchmarked using data from a currently restructured electricity market.
In this research, the output of an arrayed type tactile sensor is simulated by using a finite element method. The information of this output is processed using a neural network, and the 14 types of objects are recogni...
详细信息
In this research, the output of an arrayed type tactile sensor is simulated by using a finite element method. The information of this output is processed using a neural network, and the 14 types of objects are recognized. This recognition succeeded even if the contact position and the rotation angle of these objects are changed.
暂无评论