China
Africa
Explore chapters and articles related to this topic
Physics-Infused Learning: A DNN and GAN Approach
Published in Anuj Karpatne, Ramakrishnan Kannan, Vipin Kumar, Knowledge-Guided Machine Learning, 2023
Zhibo Zhang, Ryan Nguyen, Souma Chowdhury, Rahul Rai
The generalizability refers to the ability of a learned model to fit unseen instances. Generalization error is a measure of how accurately an algorithm is able to predict outcome values for unseen test points inside the training input domain (i.e., input space spanned by the training set). The generalization error expectation is defined by mean squared error (MSE), which measures the average squared difference between the prediction value and ground truth value. The MSE can be expressed by Equation (13.20): MSE=∑i=1n(y^i−yi)2n
Swarm Intelligence Based Algorithm for Feature Selection in High-Dimensional Datasets
Published in Wellington Pinheiro dos Santos, Juliana Carneiro Gomes, Valter Augusto de Freitas Barbosa, Swarm Intelligence Trends and Applications, 2023
Nandini Nayar, Sachin Ahuja, Shaily Jainl
The objective of feature selection is to choose a minimal feature subset, i.e., the best feature subset comprising of k-features yielding the least generalization errors. It is expected that feature selection techniques are utilized either as a pre-processing step or in combination with the learning model for the task of classification. Usually, the performance of the model is estimated in terms of the “classification rate” that is obtained on a testing set. The set of all original features are given as an input to feature selection methods, which will subsequently generate “feature subsets”. Then, the subset which is selected is evaluated by the learning algorithm or through consideration of data characteristics.
Performance and Feasibility Model Generation Using Learning-Based Approach
Published in Soumya Pandit, Chittaranjan Mandal, Amit Patra, Nano-Scale CMOS Analog Circuits, 2018
Soumya Pandit, Chittaranjan Mandal, Amit Patra
Optimal values of the hyperparameters are usually determined by minimizing the estimated generalization error. The generalization error is a function that measures the generalization ability of the constructed models, i.e., the ability to predict correctly the performance of an unknown sample. The two commonly used techniques for estimating the generalization error are [53]: Hold-out method: This is a simple technique for estimating the generalization error. The dataset is separated into two sets, called the training set and the test set. The SVM is constructed using the training set only. Then it is tested using the test dataset. The test data are completely unknown to the estimator. The mean test error which is computed by considering the errors over all test samples is used to evaluate the model. This method is very fast. However, its evaluation can have a high variance. The evaluation depends heavily on the data points that end up in the training set and on those which end up in the test set. Consequently the evaluation may be significantly different depending on how the division is made.κ-fold cross validation method: In this method, the training data is randomly split into κ mutually exclusive subsets (the folds) of approximately equal size [53]. The SVM is constructed using κ – 1 of the subsets and then tested on the subset left out. This procedure is repeated κ times. An estimation of the expected generalization error is obtained by averaging the test error over the κ trials. The advantage of this method is that the accuracy of the constructed SVM does not depend upon the division of data. The variance of the resulting estimate is reduced as κ is increased. The disadvantage of this method is that it is time consuming.
Multi-task Gaussian process upper confidence bound for hyperparameter tuning and its application for simulation studies of additive manufacturing
Published in IISE Transactions, 2023
Bo Shen, Raghav Gnanasambandam, Rongxuan Wang, Zhenyu James Kong
Different from single-task BO introduced above, Multi-Task Bayesian Optimization (MTBO) (Swersky et al., 2013) is a general method to efficiently optimize multiple different but correlated “black-box” functions. The settings for multi-task Bayesian optimization widely exist in many real-world applications. For example, the K-fold cross-validation (Bengio and Grandvalet, 2004) is a widely used technique to estimate the generalization error of a machine learning model for a given set of hyperparameters. However, it needs to retrain a model K times using all K training-validation splits. The validation errors of a model trained on K different training-validation splits can be treated as K “black-box” functions, which need to be minimized as K different tasks. These K tasks will be highly correlated, as the data are randomly partitioned among K training-validation splits. The performance of our proposed method in the application of fast cross-validation (see Swersky et al. (2013); Moss et al. (2020)) is presented in Section 5.1, which aims at minimizing the average validation errors in K-fold cross-validation.
Travel mode choice: a data fusion model using machine learning methods and evidence from travel diary survey data
Published in Transportmetrica A: Transport Science, 2019
Ximing Chang, Jianjun Wu, Hao Liu, Xiaoyong Yan, Huijun Sun, Yunchao Qu
A variety of machine learning classifiers have their own characteristics and advantages, but the prediction for a specific problem shows its shortcomings (Hagenauer and Helbich 2017). First, these studies deal only with limited machine learning classifiers, especially for travel model choice analysis. Although these classifiers have been shown to produce high precision results, there is no comparison with the newly-emerging ensemble model with stacking strategies. Second, the model does not consider the sampling differences, with poor performance on the test set, thus the generalization error is larger. In this paper, the use of cross-validation and resampling techniques lead to lower generalization errors. Third, there is a need to study the importance of attributes affecting travel mode choice and the inherent links between attributes, which is the foundation of model construction (Murray and Conner 2009).
Green Progress of Cross-border E-Commerce Industry Utilizing Random Forest Algorithm and Panel Tobit Model
Published in Applied Artificial Intelligence, 2023
The generalization error can be estimated using techniques such as cross-validation, where the model is trained and tested on different subsets of the available data. The generalization error is an important metric in machine learning, as it indicates the extent to which a model has overfit the training data, and therefore how well it is likely to perform on new data.