Nonlinear Kernel Research Articles

Prediction of fishbone linear instability in tokamaks with machine learning methods

Abstract A machine learning based surrogate model for fishbone linear instability in tokamaks is constructed. Hybrid simulations with the kinetic-magnetohydrodynamic (MHD) code M3D-K is used to generate the database of fishbone linear instability, through scanning the four key parameters which are thought to determine the fishbone physics. The four key parameters include (1) central total beta of both thermal plasma and fast ions, (2) the fast ion pressure fraction, (3) central value of safety factor q and (4) the radius of q = 1 surface. Four machine learning methods including linear regression, support vector machines (SVM) with linear kernel, SVM with nonlinear kernel and multilayer perceptron are used to predict the fishbone instability, growth rate and real frequency, mode structure respectively. Among the four methods, SVM with nonlinear kernel performs very well to predict the linear instability with accuracy ≈ 95%, growth rate and real frequency with R2 ≈ 98%, mode structure with R2 ≈ 98%.

Random Fourier features based nonlinear recurrent kernel normalized LMS algorithm with multiple feedbacks

The performance of kernel adaptive filtering algorithms (KAFs) with nonlinear recurrent structures surpasses traditional KAFs, attributed to the nonlinear contribution of feedback. Nevertheless, the existing nonlinear recurrent KAFs primarily focus on a single feedback output, potentially limiting their latent filtering capabilities. In this paper, we introduce a novel alternative, named nonlinear recurrent kernel normalized least-mean-square with multiple feedbacks (NR-KNLMS-MF), which leverages the information from multiple feedback outputs. Additionally, to tackle the computational complexity challenges associated with KAFs, we integrate random Fourier features (RFF) into NR-KNLMS-MF, resulting in an efficient variant called as RFF-NR-KNLMS-MF. Furthermore, we conduct a theoretical analysis of the mean-square convergence for RFF-NR-KNLMS-MF. Simulation results on time-series predictions demonstrate the superiority of our proposed algorithms over other competing alternatives, validating their effectiveness.

Extraordinarily Time- and Memory-Efficient Large-Scale Canonical Correlation Analysis in Fourier Domain: From Shallow to Deep.

Canonical correlation analysis (CCA) is a correlation analysis technique that is widely used in statistics and the machine-learning community. However, the high complexity involved in the training process lays a heavy burden on the processing units and memory system, making CCA nearly impractical in large-scale data. To overcome this issue, a novel CCA method that tries to carry out analysis on the dataset in the Fourier domain is developed in this article. Appling Fourier transform on the data, we can convert the traditional eigenvector computation of CCA into finding some predefined discriminative Fourier bases that can be learned with only element-wise dot product and sum operations, without complex time-consuming calculations. As the eigenvalues come from the sum of individual sample products, they can be estimated in parallel. Besides, thanks to the data characteristic of pattern repeatability, the eigenvalues can be well estimated with partial samples. Accordingly, a progressive estimate scheme is proposed, in which the eigenvalues are estimated through feeding data batch by batch until the eigenvalues sequence is stable in order. As a result, the proposed method shows its characteristics of extraordinarily fast and memory efficiencies. Furthermore, we extend this idea to the nonlinear kernel and deep models and obtained satisfactory accuracy and extremely fast training time consumption as expected. An extensive discussion on the fast Fourier transform (FFT)-CCA is made in terms of time and memory efficiencies. Experimental results on several large-scale correlation datasets, such as MNIST8M, X-RAY MICROBEAM SPEECH, and Twitter Users Data, demonstrate the superiority of the proposed algorithm over state-of-the-art (SOTA) large-scale CCA methods, as our proposed method achieves almost same accuracy with the training time of our proposed method being 1000 times faster. This makes our proposed models best practice models for dealing with large-scale correlation datasets. The source code is available at https://github.com/Mrxuzhao/FFTCCA.

Sub-sampling graph neural networks for genomic prediction of quantitative phenotypes.

In genomics, use of deep learning (DL) is rapidly growing and DL has successfully demonstrated its ability to uncover complex relationships in large biological and biomedical data sets. With the development of high-throughput sequencing techniques, genomic markers can now be allocated to large sections of a genome. By analysing allele sharing between individuals, one may calculate realized genomic relationships from single nucleotide polymorphisms (SNPs) data rather than relying on known pedigree relationships under polygenic model. The traditional approaches in genome-wide prediction (GWP) of quantitative phenotypes utilise genomic relationships in fixed global covariance modelling, possibly with some non-linear kernel mapping (for example Gaussian processes). On the other hand, the DL approaches proposed so far for GWP fail to take into account the non-Euclidean graph structure of relationships between individuals over several generations. In this paper, we propose one global convolutional neural network (GCN) and one local sub-sampling architecture (GCN-RS) that are specifically designed to perform regression analysis based on genomic relationship information. A GCN is tailored to non-Euclidean spaces and consists of several layers of graph convolutions. The GCN-RS architecture is designed to further improve the GCN's performance by sub-sampling the graph to reduce the dimensionality of the input data. Through these graph convolutional layers, the GCN maps input genomic markers to their quantitative phenotype values. The graphs are constructed using an iterative nearest neighbour approach. Comparisons show that the GCN-RS outperforms the popular Genomic Best Linear Unbiased Predictor (GBLUP) method on one simulated and three real data sets from wheat, mice and pig with a predictive improvement of 4.4% to 49.4% in terms of test mean squared error (MSE). This indicates that GCN-RS is a promising tool for genomic predictions in plants and animals. Furthermore, GCN-RS is computationally efficient, making it a viable option for large-scale applications.

Thermomechanical interaction in a living tissue due to variable thermal loading with memory

AbstractIn order to address various clinical applications within living tissue, the aim of this work is to analytically study the thermomechanical interaction for a living tissue which is subjected to variable thermal loadings. Human tissues undergoing regional hyperthermia treatment for cancer therapy is based on graded changes of the cells, and as a consequences, the constitutive equations have been formulated using the nonlocal elasticity theory. The heat transport equation for the present problem is formulated in the context of Moore‐Gibson‐Thompson theory of generalized thermoelasticity assimilating the memory‐dependent derivative within a slipping interval. Both the boundaries of the tissue is maintaining the condition of zero traction. The lower boundary of the tissue is subjected to prescribed thermal loading while, the upper boundary is kept at zero temperature. Utilizing the Laplace transform mechanism, the governing equations have been solved and the general solutions have been obtained in the transformed domain. In order to arrive at the solutions in the real space‐time domain, suitable inversion of the Laplace transform has been carried out numerically using the method of Zakian. Numerical findings suggest that thermomechanical waves propagate through skin tissue over finite distances, which helps mitigate the unrealistic predictions made by the Pennes' model. Significant effect due to different effective parameter such as nonlocal parameter and the time‐delay parameter is reported. Also, how a nonlinear kernel function can be more effective in bio‐heat transfer, is outlined in the study also.

Resting-state frontal electroencephalography (EEG) biomarkers for detecting the severity of chronic neuropathic pain

Increasing evidence is present to enable pain measurement by using frontal channel EEG-based signals with spectral analysis and phase-amplitude coupling. To identify frontal channel EEG-based biomarkers for quantifying pain severity, we investigated band-power features to more complex features and employed various machine learning algorithms to assess the viability of these features. We utilized a public EEG dataset obtained from 36 patients with chronic pain during an eyes-open resting state and performed correlation analysis between clinically labelled pain scores and EEG features from Fp1 and Fp2 channels (EEG band-powers, phase-amplitude couplings (PAC), and its asymmetry features). We also conducted regression analysis with various machine learning models to predict patients’ pain intensity. All the possible feature sets combined with five machine learning models (Linear Regression, random forest and support vector regression with linear, non-linear and polynomial kernels) were intensively checked, and regression performances were measured by adjusted R-squared value. We found significant correlations between beta power asymmetry (r = −0.375), gamma power asymmetry (r = −0.433) and low beta to low gamma coupling (r = −0.397) with pain scores while band power features did not show meaningful results. In the regression analysis, Support Vector Regression with a polynomial kernel showed the best performance (R squared value = 0.655), enabling the regression of pain intensity within a clinically usable error range. We identified the four most selected features (gamma power asymmetry, PAC asymmetry of theta to low gamma, low beta to low/high gamma). This study addressed the importance of complex features such as asymmetry and phase-amplitude coupling in pain research and demonstrated the feasibility of objectively observing pain intensity using the frontal channel-based EEG, that are clinically crucial for early intervention.

Bayesian inference of the spatial distribution of steel corrosion in reinforced concrete structures using corrosion-induced crack width

Observations of corrosion-induced crack widths offer crucial information about the corrosion states of steel reinforcements in reinforced concrete (RC) structures, enabling a cost-effective method for inferring corrosion states through inverse analysis. However, the uncertainty associated with the relationship between corrosion-induced cracking and steel weight loss necessitates a probabilistic inference method, especially when considering the spatial distributions of steel weight loss, which provides important information to estimate the load-bearing capacity loss of corroded RC structures. This paper proposes a Bayesian framework to infer the steel weight loss distribution in RC structures based on the observed corrosion-induced crack width. To reduce the dimensions of the Bayesian inference, a Karhunen-Loève transform is applied to extract the principal distribution features of the steel weight loss. The forward model of the Bayesian inference adopts a data-driven sequence-to-sequence transduction approach to predict corrosion-induced crack width from steel weight loss. This model incorporates a novel nonlinear convolution kernel for input encoding and a sparse polynomial chaos expansion for decoding, which proves more accurate and efficient than finite element simulations. The Hamiltonian Markov chain Monte Carlo (HMCMC) sampler is used to efficiently sample from the posterior distribution of the Bayesian inference. The case study of the proposed method demonstrated that Bayesian inference provides robust range estimation for the steel weight loss distribution, with its 95% confidence interval encompassing most observations. Additionally, the method efficiently inferred high-dimensional steel weight loss sequences up to 61 dimensions, taking advantage of the dimension reduction technique and the gradient-informed HMCMC sampler.

Adaptive Sparse Regular Split Gaussian Kernel Least Mean Square Algorithm for Super-Low-Frequency Motion-Induced Noise Cancellation

In super-low-frequency (SLF) submarine communication, the motion-induced noise of the towed antenna is the primary noise source, and below 500 Hz, it increases with increasing speed. We propose an improved quadratic Approximate Forward–Backward Split Gaussian Kernel Least Mean Square Algorithm (ASRSG–KLMS) based on the forward–backward split criterion using noise approximation of the nonlinear kernel least mean square, which introduces an L2-paradigm regularization term and has good sparsity while maintaining optimization stability. The ASRSG–KLMS algorithm could improve the narrowband signal-to-noise ratio by approximately 6.93 dB in the frequency range of 45–55 Hz, making it suitable for motion-induced noise cancellation in the SLF band.

Effectiveness of nonlinear kernel with memory for a functionally graded solid with size dependency

Effectiveness of nonlinear kernel with memory for a functionally graded solid with size dependency

Comparative Temporal Analysis of SVM-Based Machine Learning Techniques for Bank Risk Assessment

This paper introduces an innovative approach that utilizes support vector machines (SVMs) to predict the bankruptcy of financial institutions within the United States. The study aims to identify influential factors that contributed to bankruptcy during two significant peri-ods: the 2008 financial crisis and post-2013. The goal is to highlight both shared and distinc-tive characteristics between these time frames. The proposed method incorporates a meticu-lous feature selection procedure to identify the most critical variables for assessing a banks financial stability. Subsequently, the SVM model is fed with data containing these key vari-ables from various banks, initiating both the training and testing phases. Specifically, two SVM models were trained: one utilizing a linear kernel, and the other employing a non-linear kernel. The objective was to assess their effectiveness in distinguishing between solvent and insolvent banks. Moreover, a neural network model was developed and subjected to a com-parative analysis alongside the aforementioned SVM models, all with the aim of identifying the optimal method for bankruptcy prediction. The training dataset comprised data from the ten quarters preceding bank failures post-2013, as well as the eight quarters leading up to bank failures in 2010, during 2008 financial crisis. The SVM models were implemented us-ing Scikit-Learn, while the neural network model was trained using PyTorch. Through this comprehensive approach, the paper contributes to the advancement of predictive methodolo-gies for identifying potential financial institution bankruptcies.

Existence results, regularity and compactness properties, in the $$\alpha $$-norm, for semilinear partial functional integrodifferential equations with nonlinear Kernel and delay argument

Existence results, regularity and compactness properties, in the $$\alpha $$-norm, for semilinear partial functional integrodifferential equations with nonlinear Kernel and delay argument

Intuitionistic fuzzy twin proximal SVM with fuzzy hyperplane and its application in EEG signal classification

Twin support vector machine (TSVM) is a contemporary machine learning technique to tackle classification and regression problems. However, TSVM lacks the ability to differentiate between support vectors and noises since it neglects the positional information of input data samples, and hence it is sensitive to noises. Additionally, it fails to consider uncertainties associated with the data, which reduces its capacity for generalization. To address these drawbacks, we propose a novel fuzzy hyperplane based intuitionistic fuzzy twin proximal support vector machine. The first significant feature of the proposed approach is that it gives an intuitionistic fuzzy number based on the relevance to each data vector. This efficiently reduces the impact of noise and outliers by leveraging the local neighborhood information within the data points by incorporating membership and non-membership weights. Secondly, all the parameters present in the model are fuzzy variables, including the offset term and the elements of the normal vector. The suggested fuzzy hyperplane successfully reflects the inherent ambiguity prevalent in real-world categorization problems by reflecting vagueness in the input data through the use of fuzzy variables. The model’s efficiency is enhanced by solving two systems of linear equations to obtain two non-parallel classifiers rather than solving two quadratic programming problems as in standard TSVM. Utilizing non-linear kernel functions within the feature space enables the method to effectively identify complex patterns or non-linear relationships within the datasets. In order to demonstrate the effectiveness of the suggested approach, extensive computer experiments have been conducted on a set of eighteen benchmark datasets with both linear and non-linear kernels. In addition, rigorous statistical analysis, including Friedman and post-hoc Nemenyi tests, have been employed to assess the significance of the observed performance differences. Furthermore, we also performed numerical experiments utilizing linear, Gaussian, and polynomial kernels to classify electroencephalogram (EEG) signals. The outcomes of the experiment are analyzed in terms of average accuracy, processing time, and F-measure. The results demonstrate that the proposed method outperforms existing methods and achieves better generalization.

Using Data Science Tools to Reveal and Understand Subtle Relationships of Inhibitor Structure in Frontal Ring-Opening Metathesis Polymerization.

The rate of frontal ring-opening metathesis polymerization (FROMP) using the Grubbs generation II catalyst is impacted by both the concentration and choice of monomers and inhibitors, usually organophosphorus derivatives. Herein we report a data-science-driven workflow to evaluate how these factors impact both the rate of FROMP and how long the formulation of the mixture is stable (pot life). Using this workflow, we built a classification model using a single-node decision tree to determine how a simple phosphine structural descriptor (Vbur-near) can bin long versus short pot life. Additionally, we applied a nonlinear kernel ridge regression model to predict how the inhibitor and selection/concentration of comonomers impact the FROMP rate. The analysis provides selection criteria for material network structures that span from highly cross-linked thermosets to non-cross-linked thermoplastics as well as degradable and nondegradable materials.

Connectome embedding in multidimensional graph spaces

Abstract Connectomes' topological organization can be quantified using graph theory. Here, we investigated brain networks in higher dimensional spaces defined by up to ten graph-theoretic nodal properties. These properties assign a score to nodes, reflecting their meaning in the network. Using 100 healthy unrelated subjects from the Human Connectome Project, we generated various connectomes (structural/functional, binary/weighted). We observed that nodal properties are correlated (i.e., they carry similar information) at whole-brain and subnetwork level. We conducted an exploratory machine learning analysis to test whether high dimensional network information differs between sensory and association areas. Brain regions of sensory and association networks were classified with an 80-86% accuracy in a 10D space. We observed the largest gain in machine learning accuracy going from a 2D to 3D space, with a plateauing accuracy towards 10D space, and nonlinear Gaussian kernels outperformed linear kernels. Finally, we quantified the Euclidean distance between nodes in a 10D graph space. The multidimensional Euclidean distance was highest across subjects in the default mode network (in structural networks) and frontoparietal and temporal lobe areas (in functional networks). To conclude, we propose a new framework for quantifying network features in high-dimensional spaces that may reveal new network properties of the brain.

SAR target recognition through adaptive kernel sparse representation model based on local contrast perception

Sparse representation-based methods have shown promising prospects in synthetic aperture radar (SAR) target recognition due to their robustness under complex application conditions. However, these methods are still susceptible to the feature representation of SAR images. In this paper, we propose a new SAR target recognition method using an adaptive kernel sparse representation model based on local contrast perception (LCP). The LCP method combines a specially designed random projection matrix with two-dimensional filtering to capture extensive intensity contrasts between local regions in SAR images. A response map reconstruction technique shows that LCP can obtain discriminative properties between target and background regions and improve feature representation. The LCP features are then transformed into the kernel representation to improve linear separability by defining a reproducing kernel Hilbert space (RKHS) and using a nonlinear kernel function. The kernel representation is utilized to construct a new kernel sparse representation model for classification. Furthermore, neighborhood analysis (NA) is proposed to establish the relationship between the parameter and feature distribution, with an adaptive regularization parameter selection strategy to mitigate the parameter sensitivity. Experiments conducted on the moving and stationary target acquisition and recognition dataset demonstrate the superiority and robustness of the proposed method under different application conditions.

Accounting for nonlinear responses to traits improves range shift predictions

AbstractAccurately predicting species' range shifts in response to environmental change is paramount for understanding ecological processes and global change. In synthetic analyses, traits emerge as significant but weak predictors of species' range shifts across recent climate change. These studies assume linear responses to traits, while detailed empirical work often reveals trait responses that are unimodal and contain thresholds or other nonlinearities. We hypothesize that the use of linear modeling approaches fails to capture these nonlinearities and, therefore, may be under‐powering traits to predict range shifts. We evaluate the predictive performance of approaches that can capture nonlinear relationships (ridge‐regularized linear regression, support vector regression with linear and nonlinear kernels, and random forests). We apply our models using six multidecadal range shift datasets for plants, moths, marine fish, birds, and small mammals. We show that nonlinear approaches can perform better than least‐squares linear modeling in reproducing historical range shifts. Consistent with expectations, we identify dispersal and climatic niche traits as primary determinants of distribution shifts. Traits identified as important predictors and the direction of trait effects are generally consistent across models, but there are notable exceptions. Among important predictors, there are more consistent responses to climatic niches than dispersal ability. Modest improvements in predictability when accounting for nonlinearities and interactions, and the overall low amount of variance accounted for by trait predictors suggest limits to trait‐based statistical predictive frameworks.

“Using network analysis modularity to group health code systems and decrease dimensionality in machine learning models”

BackgroundMachine learning (ML) prediction models in healthcare and pharmacy-related research face challenges with encoding high-dimensional Healthcare Coding Systems (HCSs) such as ICD, ATC, and DRG codes, given the trade-off between reducing model dimensionality and minimizing information loss. ObjectivesTo investigate using Network Analysis modularity as a method to group HCSs to improve encoding in ML models. MethodsThe MIMIC-III dataset was utilized to create a multimorbidity network in which ICD-9 codes are the nodes and the edges are the number of patients sharing the same ICD-9 code pairs. A modularity detection algorithm was applied using different resolution thresholds to generate 6 sets of modules. The impact of four grouping strategies on the performance of predicting 90-day Intensive Care Unit readmissions was assessed. The grouping strategies compared: 1) binary encoding of codes, 2) encoding codes grouped by network modules, 3) grouping codes to the highest level of ICD-9 hierarchy, and 4) grouping using the single-level Clinical Classification Software (CCS). The same methodology was also applied to encode DRG codes but limiting the comparison to a single modularity threshold to binary encoding.The performance was assessed using Logistic Regression, Support Vector Machine with a non-linear kernel, and Gradient Boosting Machines algorithms. Accuracy, Precision, Recall, AUC, and F1-score with 95% confidence intervals were reported. ResultsModels utilized modularity encoding outperformed ungrouped codes binary encoding models. The accuracy improved across all algorithms ranging from 0.736 to 0.78 for the modularity encoding, to 0.727 to 0.779 for binary encoding. AUC, recall, and precision also improved across almost all algorithms. In comparison with other grouping approaches, modularity encoding generally showed slightly higher performance in AUC, ranging from 0.813 to 0.837, and precision, ranging from 0.752 to 0.782. ConclusionsModularity encoding enhances the performance of ML models in pharmacy research by effectively reducing dimensionality and retaining necessary information. Across the three algorithms used, models utilizing modularity encoding showed superior or comparable performance to other encoding approaches. Modularity encoding introduces other advantages such as it can be used for both hierarchical and non-hierarchical HCSs, the approach is clinically relevant, and can enhance ML models' clinical interpretation. A Python package has been developed to facilitate the use of the approach for future research.

Non-linear Kernel Optimisation of Support Vector Machine Algorithm for Online Marketplace Sentiment Analysis

Twitter is a social media platform that is very important in the digital world. Fast communication and interaction make Twitter a vital information center in sentiment analysis. The purpose of this research is to classify public opinion about the presence of marketplaces in Indonesia, both positive and negative sentiments, using a Non-linear SVM algorithm based on 1276 tweets. This research involves the stages of data pre-processing, labeling, feature extraction using TF-IDF, and data division into three scenarios: 80% training data and 20% test data, 50% training data and 50% test data scenario, and 20% training data and 80% test data scenario. The last process, GridSearchCV, combines cross-validation and non-linear SVM parameters for model evaluation using a confusion matrix. The best SVM model resulting from the scenario was 80% training and 20% test data, with hyperparameters Gamma = 100 and C = 0.01, achieving 89% accuracy. When tested on never-before-seen data, the accuracy increased to 90%, with an f1-score of 91%, precision of 88%, and recall of 95% on negative sentiments. In conclusion, evaluating the performance of non-linear SVM kernels with a combination of hyperparameter values can improve accuracy, especially on public response information about online marketplaces and public sentiment.

Fast, accurate, and interpretable decoding of electrocorticographic signals using dynamic mode decomposition

Dynamic mode (DM) decomposition decomposes spatiotemporal signals into basic oscillatory components (DMs). DMs can improve the accuracy of neural decoding when used with the nonlinear Grassmann kernel, compared to conventional power features. However, such kernel-based machine learning algorithms have three limitations: large computational time preventing real-time application, incompatibility with non-kernel algorithms, and low interpretability. Here, we propose a mapping function corresponding to the Grassmann kernel that explicitly transforms DMs into spatial DM (sDM) features, which can be used in any machine learning algorithm. Using electrocorticographic signals recorded during various movement and visual perception tasks, the sDM features were shown to improve the decoding accuracy and computational time compared to conventional methods. Furthermore, the components of the sDM features informative for decoding showed similar characteristics to the high-γ power of the signals, but with higher trial-to-trial reproducibility. The proposed sDM features enable fast, accurate, and interpretable neural decoding.

NorSand4AI: a comprehensive triaxial test simulation database for NorSand constitutive model materials

Abstract. In soil sciences, parametric models known as constitutive models (e.g., the Modified Cam Clay and the NorSand) are used to represent the behavior of natural and artificial materials. In contexts where liquefaction may occur, the NorSand constitutive model has been extensively applied by both industry and academia due to its relatively simple critical state formulation and low number of input parameters. Despite its suitability as a good modeling framework to assess static liquefaction, the NorSand model still is based on premises which may not perfectly represent the behavior of all soil types. In this context, the creation of data-driven and physically informed metamodels emerges. The literature suggests that data-driven models should initially be developed using synthetic datasets to establish a general framework, which can later be applied to experimental datasets to enhance the model's robustness and aid in discovering potential mechanisms of soil behavior. Therefore, creating large and reliable synthetic datasets is a crucial step in constructing data-driven constitutive models. In this context, the NorSand model comes in handy: by using NorSand simulations as the training dataset, data-driven constitutive metamodels can then be fine-tuned using real test results. The models created that way will combine the power of NorSand with the flexibility provided by data-driven approaches, enhancing the modeling capabilities for liquefaction. Therefore, for a material following the NorSand model, the present paper presents a first-of-its-kind database that addresses the size and complexity issues of creating synthetic datasets for nonlinear constitutive modeling of soils by simulating both drained and undrained triaxial tests. Two datasets are provided: the first one considers a nested Latin hypercube sampling of input parameters encompassing 2000 soil types, each subjected to 40 initial test configurations, resulting in a total of 160 000 triaxial test results. The second one considers nested quasi-Monte Carlo sampling techniques (Sobol and Halton) of input parameters encompassing 2048 soil types, each subjected to 42 initial test configurations, resulting in a total of 172 032 triaxial test results. By using the quasi-Monte Carlo dataset and 49 of its subsamples, it is shown that the dataset of 2000 soil types and 40 initial test configurations is sufficient to represent the general behavior of the NorSand model. In this process, four machine learning algorithms (Ridge Regressor, KNeighbors Regressor and two variants of the Ridge Regressor which incorporate nonlinear Nystroem kernel mappings of the input and output values) were trained to predict the constitutive and test parameters based solely on the triaxial test results. These algorithms achieved 13.91 % and 16.18 % mean absolute percentage errors among all 14 predicted parameters for undrained and drained triaxial test inputs, respectively. As a secondary outcome, this work introduces a Python script that links the established Visual Basic implementation of NorSand to the Python environment. This enables researchers to leverage the comprehensive capabilities of Python packages in their analyses related to this constitutive model.