Linear Regression Method for Review Aspect Rating Prediction

Online reviews as a new textual domain offer a unique proposition for sentiment analysis. The reviewers usually give a whole rating score to the product. The potential customers tend to make decision according to the reviews. Previous works mainly focus on the summarization of the rating and sentiment of reviews. However, they ignore an important question. The whole rating can be regarded as linear regression of different aspect ratings. High aspect rating and low aspect rating compensate each other. Therefore, previous works are coarse-grained analysis. This paper first proposed a weak supervised learning method to extract implicit aspect with aspect seeds. It then formulates the aspect rating problem as a linear regression model. Finally a gradient descent method is proposed to handle the problem. Different datasets are collected. Experimental result in the datasets demonstrates the advantage of the proposed model.

Download Full-text

Search Strategy for Rare Microstructure to Optimize Material Properties of Filled Rubber

10.21203/rs.3.rs-144773/v1 ◽

2021 ◽

Author(s):

Takashi Kojima ◽

Takashi Washio ◽

Satoshi Hara ◽

Masakata Koishi

Keyword(s):

Gradient Descent ◽

Search Strategy ◽

Dimensional Space ◽

Descent Method ◽

Coarse Grained ◽

Replica Exchange ◽

Second Step ◽

Gradient Descent Method ◽

Filled Rubber ◽

Wide Range

Abstract A shortcut to understand the microstructure-property relationship is sampling and analysis of microstructures that induce the desired material property. In the case of filled rubber, the simulation of complex filler morphology involves hundreds of filler particles. This requires a large amount of iterative sampling, because the number of parameters is when using coordinates of the n particles as the search objective. Furthermore, the morphology that induces the desired property, e.g. extremely high modulus, only occurs rarely. In this paper, we propose an effective three-step search method for the filler morphology. In the first step, the replica exchange Markov chain Monte Carlo (MCMC) was employed to discretely search among a wide range of morphologies. In this step, we reduced the filler morphology space in sampling by introducing distributed filler candidate points and spin function. In the second step, the gradient descent method was applied to search for the desired morphology locally in the high-dimensional space , starting from the morphologies obtained by the replica exchange MCMC. Lastly, the coarse-grained molecular dynamics (CGMD) simulations were performed to validate the morphologies actually show the desired properties, because the surrogate model of CGMD was employed in the first 2 steps for the efficient search. Using the proposed method, we demonstrate the search for morphologies that induce high elastic modulus.

Download Full-text

SIRMs (Single Input Rule Modules) Connected Fuzzy Inference Model

Journal of Advanced Computational Intelligence and Intelligent Informatics ◽

10.20965/jaciii.1997.p0023 ◽

1997 ◽

Vol 1 (1) ◽

pp. 23-30 ◽

Cited By ~ 24

Author(s):

Naoyoshi Yubazaki ◽

◽

Jianqiang Yi ◽

Kaoru Hirota ◽

Keyword(s):

Gradient Descent ◽

Fuzzy Inference ◽

Fuzzy Rule ◽

Descent Method ◽

Gradient Descent Method ◽

Control Performance ◽

Nonlinear Functions ◽

Inference Model ◽

Proposed Model ◽

Single Input

A new fuzzy inference model, SIRMs (Single Input Rule Modules) Connected Fuzzy Inference Model, is proposed for plural input fuzzy control. For each input item, an importance degree is defined and single input fuzzy rule module is constructed. The importance degrees control the roles of the input items in systems. The model output is obtained by the summation of the products of the importance degree and the fuzzy inference result of each SIRM. The proposed model needs both very few rules and parameters, and the rules can be designed much easier. The new model is first applied to typical secondorder lag systems. The simulation results show that the proposed model can largely improve the control performance compared with that of the conventional fuzzy inference model. The tuning algorithm is then given based on the gradient descent method and used to adjust the parameters of the proposed model for identifying 4-input 1-output nonlinear functions. The identification results indicate that the proposed model also has the ability to identify nonlinear systems.

Download Full-text

Wireless Brain Wave Classification for Alzheimer’s Patients via Efficient Neural Network Computation

Advances in Data Science and Adaptive Analysis ◽

10.1142/s2424922x18500043 ◽

2018 ◽

Vol 10 (03) ◽

pp. 1850004

Author(s):

Grant Sheen

Keyword(s):

Neural Network ◽

Gradient Descent ◽

Descent Method ◽

Stochastic Gradient Descent ◽

Normal Person ◽

Gradient Descent Method ◽

Step Size ◽

Brain Wave ◽

Proposed Model ◽

Wireless Recording

Wireless recording and real time classification of brain waves are essential steps towards future wearable devices to assist Alzheimer’s patients in conveying their thoughts. This work is concerned with efficient computation of a dimension-reduced neural network (NN) model on Alzheimer’s patient data recorded by a wireless headset. Due to much fewer sensors in wireless recording than the number of electrodes in a traditional wired cap and shorter attention span of an Alzheimer’s patient than a normal person, the data is much more restrictive than is typical in neural robotics and mind-controlled games. To overcome this challenge, an alternating minimization (AM) method is developed for network training. AM minimizes a nonsmooth and nonconvex objective function one variable at a time while fixing the rest. The sub-problem for each variable is piecewise convex with a finite number of minima. The overall iterative AM method is descending and free of step size (learning parameter) in the standard gradient descent method. The proposed model, trained by the AM method, significantly outperforms the standard NN model trained by the stochastic gradient descent method in classifying four daily thoughts, reaching accuracies around 90% for Alzheimer’s patient. Curved decision boundaries of the proposed model with multiple hidden neurons are found analytically to establish the nonlinear nature of the classification.

Download Full-text