Fundamentals and Methods of Machine and Deep Learning. Pradeep Singh

Чтение книги онлайн.

СКАЧАТЬ 2.2 A high-level representation of Bootstrap aggregating.

      Some of the advantages offered by bootstrap in diagnosing the zonotic diseases are as follows: the aggregated power of several weak learners over runs the performance of strong learner, the variance incurred gets reduced as it efficiently handles overfitting problem, no loss of precision during interoperability, computationally not expensive due to proper management of resources, computation of over confidence bias becomes easier, equal weights are assigned to models which increases the performance, misclassification of samples is less, very much robust to the effect of the outliers and noise, the models can be easily paralyzed, achieves high accuracy through incremental development of the model, stabilizes the unstable methods, easier from implementation point of view, provision for unbiased estimate of test errors, easily overcomes the pitfalls of individuals machine learning models, and so on.

2.4 Bayesian Model Averaging (BMA)

It is one of the popularly referred ensemble machine learning model which applies Bayesian inference to solve the issues related to the selection of problem statement, performing the combined estimation, and produces the results using any of the straight model with less prediction accuracy. Several coherent models are available in BMA which are capable of handling the uncertainty available in the large datasets. The steps followed while implemented the MBA model is managing the summation, computation of integral values for MBA, using linear regression for predictions, and transformation purposes [18, 19].

Basically, BMA is an extended form of Bayesian inference which performs mathematical modeling of uncertainty using prior distribution by obtaining the posterior probability using Bayes theorem. For implementing the BMA, first prior distribution of each of the models in the ensemble network needs to be specified then evidence needs to be found for each of the model. Suppose the existing models are represented by Ml, where the value of l varies from 1 to k which basically represent the set of probability distributions. The probability distribution computes likelihood function L(Y|θl, Ml), where θl stands for parameter which are model specific dependent parameter. According to the Bayes theorem, the value for posterior probability is computed as follows [20]. A high-level representation of BMA is shown in Figure 2.3. Bayesian model representation begins with the data set which is distributed among multiple data subsets. Each subset of data is fed as input to the learner then average operation is performed finally compared with the average threshold and tested using permutation threshold to generate the Bayesian model as output.

      Figure 2.3 A high-level representation of Bayesian model averaging (BMA).

      Some of the advantages offered by BMA in diagnosing the zonotic diseases are as follows: capable of performing multi-variable selection, generates overconfident inferences, the number of selected features are less, easily scalable to any number of classes, posterior probability efficiency is high, deployment of the model is easier, correct estimation of uncertainty, suitable to handle complex applications, proper accounting of the model, combines estimation and predictions, flexible with prior distribution, uses mean candidate placement model, performs multi-linear operation, suitable of handling the heterogeneous resources, provides transparent interpretation of the large amount of data, error reduction happens exponentially, the variance incurred in prediction is less, flexibility achieved in parameter inference is less, prediction about model prediction is less, high-speed compilation happens, generated high valued output, combines efficiency achieved by several learner and average models, very much robust against the effect caused by misspecification of input attributes, model specification is highly dynamic, and so on.

2.5 Bayesian Classifier Combination (BCC)

      Bayesian classifier combination (BCC) considers k different types of classifiers and produces the combined output. The motivation behind the innovation of this classifier is will capture the exhaustive possibilities about all forms of data, and ease of computation of marginal likelihood relationships. This classifier will not assume that the existing classifiers are true rather it is assumed to be probabilistic which mimics the behavior of the human experts. The BCC classifier uses different confusion matrices employed over the different data points for classification purpose. If the data points are hard, then the BCC uses their own confusion matrix; else, the posterior confusion matrix will be made use. The classifier identifies the relationship between the output of the model and the unknown data labels. The probabilistic models are not required; they share information about sending or receiving the information about the training data [21, 22].

The BCC model the parameters which includes

hyperparameters

. Based on the values of the prior posterior probability distribution of random variables with observed label classes, the independence posterior density id computed as follows:

      The inferences drawn are based on the unknown random variables, i.e., P, π, t, V, and α which are collected using Gibbs and rejection sampling methodology. A high-level representation of BCC is shown in Figure 2.4. First parameters of BCC model, hyperparameters, and posterior probabilities are summed to generate final prediction as output.

Some of the advantages offered by BCC in diagnosing the zonotic diseases are as follows: performs probabilistic prediction, isolates the outliers which causes noise, efficient handling of missing values, robust handling of irrelevant attributes, side effects caused by dependency relationships can be prevented, easier in terms of implementation, ease modeling of dependency relationships among the random variables, learns collectively from labeled and unlabeled input data samples, ease feature selection, lazy learning, training time is less, eliminates unstable estimation, high knowledge is attained in terms of systems variable dependencies, high accuracy achieved in interpretation of the results, confusion matrix–based processing of data, low level of computational complexity, easily operates with less computational resources, requires less amount of training data, capable enough to handle the uncertainty in the data parameters, can learn from both labeled and unlabeled data samples, precise selection of the attributes which yields maximum information gain, eliminates the redundant values, lower number of tunning parameters, less memory requirement, highly flexible classification of data, and so on [23].

      Figure 2.4 A high-level representation of Bayesian classifier combination (BCC).

2.6 Bucket of Models

The bucket of models is one of the popular ensemble machine learning techniques used to choose the best algorithm for solving any computational intensive problems. The performance achieved by bucket of models is good compared to average of all ensemble machine learning models. One of the common strategies used to select the best model for prediction is through cross-validation. During cross-validation, all examples available in the training will be used to train the model and the best model which fits the problem will be chosen. One of the popular generalization approaches for cross-validation selection is gating. In order to implement the gating, the perceptron model will be used which assigns weight to the prediction product by each model available in the bucket. When the large number of models in the bucket is applied over a larger set of problems, the model for which the СКАЧАТЬ