Using Persistence Barcode to Show the Impact of Data Complexity on the Neural Network Architecture

Abstract

It is so much noticeable that initialization of architectural parameters has a great impact on whole learnability stream so that knowing mathematical properties of dataset results in providing neural network architecture a better expressivity and capacity. In this paper, five random samples of the Volve field dataset were taken. Then a training set was specified and the persistent homology of the dataset was calculated to show impact of data complexity on selection of multilayer perceptron regressor (MLPR) architecture. By using the proposed method that provides a well-rounded strategy to compute data complexity. Our method is a compound algorithm composed of the t-SNE method, alpha-complexity algorithm, and a persistence barcode reading method to extract the Betti number of a dataset. After that, MLPR were trained using that dataset using a single hidden layer with increased hidden neurons. Then, increased both hidden layers and hidden neurons. Our empirical analysis has shown that the training efficiency of MLPR severely depends on its architecture’s ability to express the homology of the dataset.

Optimizing Neural Network Efficiency by Neuron Pruning

“…Neurons with the lowest levels of sensitivity emerge as candidates for pruning, as they are considered less critical for the current model. Algorithm (1) shows the steps of sensitivity-based pruning Algorithm. Step 5: Pruned Neurons 𝑃𝑢𝑟𝑛𝑒𝑑 𝑁𝑒𝑡𝑤𝑜𝑟𝑘 = 𝑃𝑟𝑢𝑛𝑒𝑁𝑒𝑡𝑤𝑜𝑟𝑘(𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑁𝑒𝑡𝑤𝑜𝑟𝑘 , 𝑁𝑒𝑢𝑟𝑜𝑛 𝐾𝑒𝑒𝑝 )…”

“…Despite advancements, challenges persist in establishing optimal topologies for real-world applications, necessitating the exploration of factors such as the organization of networks, the number of neurons, hidden layers, and connections. In this pursuit, a Python code utilizing the Tensorflow library has been developed to efficiently determine the optimal number of hidden layers and neurons within a neural network [1]. Various algorithms exist for generating pruned networks, with Han et al (2015) proposing a method involving initial training, convergence assessment, and subsequent trimming.…”

Volve Oil Field S-Wave Log Data Prediction Using GBR and MLPR

“…Therefore, utilizing alternatives associated with the fourth paradigm shift of data-driven science helped offer the fossil industry a relatively cheaper and faster decision-making tool in many sectors. Artificial intelligence (AI), which is data-driven, has been widely used for data prediction in different sectors [1][2][3][4], and it has many applications within the oil and gas industry [5][6][7][8][9][10][11][12]. Shear Wave Sonic (S-wave) log data is essential for identifying the reservoir geomechanical properties [13], which is an important factor for the drilling, completion, and optimization processes, where obtaining an S-wave log requires capital investment.…”