Function Neural Network Research Articles

A novel wind speed forecasting combined model using variational mode decomposition, sparse auto-encoder and optimized fuzzy cognitive mapping network

The nonlinear, random and fluctuating characteristics of wind speed bring great challenges to its accurate forecast, so no model that can adapt to all situations. In order to solve the problem of unbalanced forecast accuracy and stability in the current wind speed forecast model, a novel and advanced wind speed combined forecast model (CFM) is proposed in this study. The CFM adopts a two-phase data processing strategy composed of variable mode decomposition-sparse autoencoder (VMD-SAE) to extract the original wind speed features, high-order fuzzy cognitive mapping (HFCM) neural network modeling and batch gradient descent optimization algorithm to make up for its shortcomings. The two-phase data processing strategy performs smoothing and feature information extraction processing on the original data. The forecast module adopts the SAE-HFCM combination strategy, and utilizes their respective advantages to achieve accurate and stable result output. The results show that this CFM has the best forecast accuracy and generalization performance compared with 7 benchmark models in datasets from three different sites. Compared with the benchmark models, the performance of CFM in point forecasting and interval forecasting, the average improvement percentage of IP_RMSE, IP_MAE and IP_MAPE minimum values are 40.9%, 40.1% and 40.6%, respectively. The performance of interval indicators is also the best, and its application prospects are broad.

Controllability and Robustness of Functional and Structural Connectomic Networks in Glioma Patients.

Previous studies suggest that the topological properties of structural and functional neural networks in glioma patients are altered beyond the tumor location. These alterations are due to the dynamic interactions with large-scale neural circuits. Understanding and describing these interactions may be an important step towards deciphering glioma disease evolution. In this study, we analyze structural and functional brain networks in terms of determining the correlation between network robustness and topological features regarding the default-mode network (DMN), comparing prognostically differing patient groups to healthy controls. We determine the driver nodes of these networks, which are receptive to outside signals, and the critical nodes as the most important elements for controllability since their removal will dramatically affect network controllability. Our results suggest that network controllability and robustness of the DMN is decreased in glioma patients. We found losses of driver and critical nodes in patients, especially in the prognostically less favorable IDH wildtype (IDHwt) patients, which might reflect lesion-induced network disintegration. On the other hand, topological shifts of driver and critical nodes, and even increases in the number of critical nodes, were observed mainly in IDH mutated (IDHmut) patients, which might relate to varying degrees of network plasticity accompanying the chronic disease course in some of the patients, depending on tumor growth dynamics. We hereby implement a novel approach for further exploring disease evolution in brain cancer under the aspects of neural network controllability and robustness in glioma patients.

Seismic Sedimentologic Study on Sequence-Sediment Evolution in Fault Basin Regions

With potential hydrocarbon source rocks, the Mesozoic Lower Cretaceous interval in the Aitegle Sag of the Yingen-Ejinaqi Basin in China has a bright petroleum exploration prospect. Considering the scarcity of wells in Aitegle Sag, we have systematically carried out the research on the sequence structure and sedimentary system evolution of the Lower Cretaceous based on seismic, core, and logging data, guided by the Vail classic sequence stratigraphy, and innovatively introduced the self-organizing mapping neural network (SOM) learning algorithm to assist in seismic geological analysis. The results are as follows: (1) after RMSAmp, LyapIndex, FrctDim, and InfoEntr clustering, five zones are divided, which are corresponding to five seismic facies (showing medium-strong amplitude and parallel-subparallel reflection; medium-strong amplitude and wedge-shaped divergent reflection; medium-strong amplitude and continuous-relatively continuous, parallel-subparallel reflection; strong amplitude and continuous sheet-like or blank reflection; and weak amplitude and mound-shaped chaotic reflection, respectively), and interpreted as five subfacies including delta plain front, prodelta-shallow lake, semideep lake, deep lake, and sublacustrine fan. (2) The Lower Cretaceous formation includes one second-order sequence, three third-order sequences (SQ2–SQ4), and six system tracts. In the SQ2–SQ4 circle, the delta system formed by sediments from southeast varies remarkably in the progradational zones in different system tracts, and tectonic movements and sediment supply control the sequence stratigraphy in the study area as major factors. (3) It is feasible to apply SOM learning algorithm to seismic attribute clustering analysis. This method effectively realizes the combination of geophysical methods based on big data analysis and geological problem analysis, which not only strengthens the use of seismic data but also ensures the effectiveness and reliability of clustering results. These findings offer basic data and scientific supports for the exploration and development of unconventional reservoirs (e.g., shale and tight oil/gas reservoirs) in the Aitegle Sag and also provide new ideas for studying the sequence-sediment in the target zones in the fault basin regions with scarce wells.

AI-Based Real-Time Star Tracker

Many systems on Earth and in space require precise orientation when observing the sky, particularly for objects that move at high speeds in space, such as satellites, spaceships, and missiles. These systems often rely on star trackers, which are devices that use star patterns to determine the orientation of the spacecraft. However, traditional star trackers are often expensive and have limitations in their accuracy and robustness. To address these challenges, this research aims to develop a high-performance and cost-effective AI-based Real-Time Star Tracker system as a basic platform for micro/nanosatellites. The system uses existing hardware, such as FPGAs and cameras, which are already part of many avionics systems, to extract line-of-sight (LOS) vectors from sky images. The algorithm implemented in this research is a “lost-in-space” algorithm that uses a self-organizing neural network map (SOM) for star pattern recognition. SOM is an unsupervised machine learning algorithm that is usually used for data visualization, clustering, and dimensionality reduction. Today’s technologies enable star-based navigation, making matching a sky image to the star map an important aspect of navigation. This research addresses the need for reliable, low-cost, and high-performance star trackers, which can accurately recognize star patterns from sky images with a success rate of about 98% in approximately 870 microseconds.

Imbalanced dataset classification using fuzzy ARTMAP and computational intelligence techniques

Recently, fuzzy adaptive resonance theory mapping (ARTMAP) neural networks are applied to solving complex problems due to their plasticity-stability capability and resonance property. An imbalanced dataset occurs when there is the presence of one class containing a greater number of instances than other classes. It is skewed representation of data. Many standard algorithms have failed in mitigating imbalanced dataset problems. There are four paradigms used-data level, algorithm level, cost-sensitive, and ensemble method in solving imbalanced dataset problems. Here we put forward a method to solve the imbalanced dataset problem by a brain-neuron framework and an ensemble of a special type of artificial neural network (ANN) called fuzzy ARTMAP thereafter we applied a clustering algorithm known as fuzzy C-means clustering to handle missing value and also propose to make fuzzy ARTMAP cost-sensitive. Results indicate 100% accuracy in classification.

Efficient SOM’s Application to Seismic Fumarolic Tremor for the Detection of Anomalous Hydrothermal Activity in Campi Flegrei Volcano (Italy)

In recent decades, the Campi Flegrei caldera (Italy) showed unrest characterized by increases in seismicity, ground uplift, and hydrothermal activity. Currently, the seismic and hydrothermal phenomena are mostly concentrated in the Solfatara–Pisciarelli area, which presents a wide fumarolic field and mud emissions. The main fumarole in Pisciarelli is associated with a boiling mud pool. Recently, episodes of a sudden increase in hydrothermal activity and expansion of mud and gas emissions occurred in this area. During these episodes, which occurred in December 2018 and September 2020, Short Duration Events (SDEs), related to the intensity of mud pool boiling, were recorded in the fumarolic seismic tremor. We applied a Self-Organizing Map (SOM) neural network to recognize the occurrence of SDEs in the fumarolic tremor of Campi Flegrei, which provides important information on the state of activity of the hydrothermal system and about the possible phreatic activity. Our method, based on an ad hoc feature extraction procedure, effectively clustered the seismic signals containing SDEs and separated them from those representing the normal fumarolic tremor. This result is useful for improving the monitoring of the Solfatara–Pisciarelli hydrothermal area which is a high-risk zone in Campi Flegrei.

Localizing Delirium through Automated Lesion Segmentation and Neural Network Mapping (N1.001)

<h3>Objective:</h3> To determine whether delirium in acute ischemic stroke is localizable to specific neural networks. <h3>Background:</h3> Delirium is estimated to occur in one-fourth of patients with acute stroke. Despite its associated morbidity and mortality, prediction and pathophysiological understanding of delirium remains challenging. A recent meta-analysis of patients with delirium and stroke found increased risks of delirium in supratentorial, cortical, and anterior circulation strokes, raising the possibility of an underlying neural network for delirium. <h3>Design/Methods:</h3> We performed a retrospective cohort study of patients admitted to a comprehensive stroke center with acute ischemic stroke from January 2016 to April 2019. Patients were assessed for delirium by trained clinical nursing staff using the Confusion Assessment Method (CAM) framework. Acute stroke lesions from MRI diffusion weighted images (DWI) lesions were automatically segmented using a machine learning algorithm. Lesions were registered to a 2mm Montreal Neurological Institute template (MNI-152). We then used lesion network mapping to identify potential unifying brain networks for delirium in patients with acute stroke. <h3>Results:</h3> 907 scans from unique patients were analyzed. 274 patients had delirium (30.2%), 47.5% were women (431), and the mean age was 69.6 years old. In comparison to a general stroke cohort, patients with delirium had lesions connected to the right and left inferior frontal gyri and left middle temporal gyrus and left temporo-parietal junction (one-sample t-test, family-wise error p&lt;0.05). Sensitivity analysis showed that delirium was not associated with parietal, occipital, brainstem, and cerebellar lesions. <h3>Conclusions:</h3> Delirium in stroke localizes to a network connected to frontotemporal regions. These results have implications for prediction of delirium in patients with acute stroke, with the potential for targeting multimodal prevention. Future work will refine their specificity, however, the results suggest that frontotemporal networks may be particularly important for cardinal symptoms of delirium, a disorder of impaired attention and awareness <b>Disclosure:</b> Dr. Rhee has stock in NTAP. Dr. Rhee has stock in TSLA. Dr. Rhee has stock in BABA. Dr. Rhee has stock in ZM. Dr. Rhee has stock in GM. Dr. Rhee has stock in PFF. Dr. Rhee has stock in AMZN. Dr. Rhee has stock in META. Dr. Rhee has stock in GE. Dr. Rhee has received publishing royalties from a publication relating to health care. Dr. Rhee has received personal compensation in the range of $0-$499 for serving as a Director of the Hippocratic Forum with Abigail Adams Institute. Dr. Ferguson has nothing to disclose. Dr. Bonkhoff has received personal compensation in the range of $500-$4,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for NeuroImage Clinical (Elsevier). Mr. Patel has nothing to disclose. Dr. Bretzner has nothing to disclose. Dr. Hong has nothing to disclose. Dr. Ryan has nothing to disclose. Dr. Westover has stock in Beacon Biosignals. The institution of Dr. Westover has received research support from NIH. Dr. Westover has received publishing royalties from a publication relating to health care. Dr. Westover has a non-compensated relationship as a cofounder with Beacon Biosignals that is relevant to AAN interests or activities. Dr. Fox has received personal compensation in the range of $500-$4,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for Wiley. Dr. Rost has received personal compensation in the range of $5,000-$9,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for Stroke - AHA/ASA Journal. The institution of Dr. Rost has received research support from NIH. Dr. Rost has received publishing royalties from a publication relating to health care. The institution of Dr. Kimchi has received research support from NIH. The institution of an immediate family member of Dr. Kimchi has received research support from NIH.

A Self-Rectifying Synaptic Memristor Array with Ultrahigh Weight Potentiation Linearity for a Self-Organizing-Map Neural Network

Two-terminal self-rectifying (SR)-synaptic memristors are preeminent candidates for high-density and efficient neuromorphic computing, especially for future three-dimensional integrated systems, which can self-suppress the sneak path current in crossbar arrays. However, SR-synaptic memristors face the critical challenges of nonlinear weight potentiation and steep depression, hindering their application in conventional artificial neural networks (ANNs). Here, a SR-synaptic memristor (Pt/NiOx/WO3-x:Ti/W) and cross-point array with sneak path current suppression features and ultrahigh-weight potentiation linearity up to 0.9997 are introduced. The image contrast enhancement and background filtering are demonstrated on the basis of the device array. Moreover, an unsupervised self-organizing map (SOM) neural network is first developed for orientation recognition with high recognition accuracy (0.98) and training efficiency and high resilience toward both noises and steep synaptic depression. These results solve the challenges of SR memristors in the conventional ANN, extending the possibilities of large-scale oxide SR-synaptic arrays for high-density, efficient, and accurate neuromorphic computing.

Dynamic changes of resting state functional network following acute ischemic stroke

Stroke, the second common cause of death in the world, is commonly considered to the well-known phenomenon of diaschisis. After stroke, regions far from the lesion can show altered neural activity. However, the comprehensive treatment recovery mechanism of acute ischemic stroke remains unclear. This study aims to investigate the impact of comprehensive treatment on resting state brain functional connectivity to reveal the therapeutic mechanism through a three time points study design. Twenty-one acute ischemic stroke patients and twenty matched healthy controls (HC) were included. Resting state functional magnetic resonance imaging (fMRI) and clinical evaluations were assessed in three stages: baseline (less than 72 h after stroke onset), post-first month and post-third month. Amplitude of low-frequency fluctuations (ALFF) and Independent component analysis (ICA) were conducted. We found: 1) stroke patients had decreased ALFF in the right cuneus (one of the important parts of the visual network). After three months, ALFF increased to the normal level; 2) the decreased functional connectivity in the right cuneus within the visual network and restored three months after onset. However, the decreased functional connectivity in the right precuneus within the default mode network restored one month after onset; 3) a significant association was found between the clinical scale score change over time and improvement in the cuneus and precuneus functional connectivity. Our results demonstrate the importance of the cuneus and precuneus within the visual network and default mode network in stroke recovery. These findings suggest that the different restored patterns of neural functional networks may contribute to the neurological function recovery. It has potential applications from stroke onset through rehabilitation because different rehabilitation phase corresponds to specific strategies.

Incorporating ecological connectivity into ecological functional zoning: A case study in the middle reaches of Yangtze River urban agglomeration

Ecological functional zoning is critical for multiobjective sustainable management of social-ecological systems. However, the framework of ecological functional zoning needs to be further improved. In terms of the index system, the existing “element-structure–function” index system mainly considers the structural characteristics that have a negative impact on ecological processes, e.g., disturbances, while ignoring positive impacts, e.g., connectivity, which may weaken the role of key nodes in the ecosystem and is not conducive to regional ecological protection. In terms of research methodology, the widely used self-organizing feature mapping neural network (SOFM) produces a number of clusters that easily exceed expectations, thus increasing management difficulties. Therefore, there is an urgent need to explore new combining methods. Here, an improved index system combining ecological connectivity and ecological functions and a traditional index system including only ecological functions were constructed, and a zoning method combining a self-organizing feature mapping neural network and fuzzy mean clustering (SOFM-FCM) was used for functional zoning of the middle reaches of the Yangtze River urban agglomeration. The advantages of functional zoning incorporating ecological connectivity were verified by comparing the results of both zoning procedures before and after improvement. The results indicate that the study area could be divided into ecological conservation areas, biodiversity conservation areas, urban development areas, and grain production areas. The zoning schemes before and after the improved framework were reliable compared with the national planning scheme, but there was 20.6% inconsistency in the results after the improved framework. This was mainly manifested by the shift from biodiversity conservation areas to ecological conservation areas and from urban development areas to food production areas. This shift was more conducive to improving spatial continuity and landscape multifunctionality, thus maximizing ecological conservation, pushing back smart urban growth, and achieving food security. This study expands the research perspective concerned with ecological functional zoning, and these results can provide an important reference for regional ecological protection and related research.

Assessment of transient thermal distribution in a moving porous plate with temperature-dependent internal heat generation using Levenberg–Marquardt backpropagation neural network

ABSTRACT In the present paper, the transient temperature distribution within a moving porous plate is scrutinized by considering the internal heat generation, convection and radiation condition. Furthermore, an effective artificial intelligence technique is proposed for discussing the transient thermal behavior in a moving plate by employing the Levenberg-Marquardt backpropagation network. Using suitable non-dimensional terms, the governing partial differential equation (PDE) is transformed into its non-dimensional form. The transformed equation is then solved using the finite difference method (FDM) and the obtained thermal values (datasets) are used as target values in the process of neural network mapping. This network uses the training, testing, and validation processes to determine the solution of several scenarios established at various process times. The behavior of dimensionless transient temperature profile has been inspected for diverse values of non-dimensional parameters such as heat generation number, radiative parameter, porosity parameter, and conductive parameter with the assistance of graphs. The significant outcomes elucidate that, the transient temperature profile of the porous plate increases with the change in time. Further, it enriches for rise in the magnitude of Peclet number and heat generation number while it drops for enhanced values of porosity parameter.

Web pages from mockup design based on convolutional neural network and class activation mapping

Web pages from mockup design based on convolutional neural network and class activation mapping

Ship hull surface reconstruction from scattered points cloud using an RBF neural network mapping technology

The ship hull surface reconstruction based on three-dimensional scattered points cloud is vital to Computer-Aided Design modeling of ship reverse engineering. There are several problems with traditional methods: the preprocessing of a large-scale scattered points cloud is complicated, and the result is susceptible to noise. Because of the “black box” characteristic, the surface reconstruction based on neural networks cannot apply to the practical engineering. To address these issues, combining the Radial Basis Function neural network and Non-Uniform Rational B-Spline interpolation algorithm, a new ship hull surface reconstruction method was proposed, which can satisfy the standard geometric model description in Computer-Aided Design system. Firstly, the Radial Basis Function neural network was used to pre-fit the three-dimensional scattered points cloud. Then the mathematical model of the surface was mapped. Finally, based on the bilinear interpolation algorithm, the mathematical model was transformed to a Non-Uniform Rational B-Spline surface to apply to the ship practical engineering. In addition, by comparing our method with the traditional method, the advantages of our method in surface reconstruction quality and surface repair ability were verified, which provided a new way for the application of Radial Basis Function neural network in ship reverse engineering.

Hardware-aware Quantization/Mapping Strategies for Compute-in-Memory Accelerators

The emerging non-volatile memory (eNVM) based mixed-signal Compute-in-Memory (CIM) accelerators are of great interest in today's AI accelerators design due to their high energy efficiency. Various CIM architectures and circuit-level designs have been proposed, showing superior hardware performance for deep neural network (DNN) acceleration. However, hardware-aware quantization strategies for CIM-based accelerators are not systematically explored. Since there are a variety of design options for neural network mapping on CIM systems while improper strategies may narrow the circuit-level design space and further limit the hardware performance, it is important to make a comprehensive early-stage design space exploration and find appropriate quantization/mapping strategies to achieve better hardware performance. In this paper, we provide a joint algorithm-hardware analysis and compare the system-level hardware performance for various design options, including quantization algorithms, data representation methods and analog-to-digital converter (ADC) configurations. This work aims to propose guidelines for choosing more hardware-friendly design options for chip architects. According to our evaluation results for CIFAR-10/100 and ImageNet classification, the properly chosen quantization approach and optimal mapping strategy (dynamic fixed-point quantization + 2’s complement representation/shifted unsigned INT representation + optimized precision ADC) could achieve ∼2 × energy efficiency and 1.2 ∼ 1.6 × throughput with 5%∼25% reduced area overhead, compared to naïve strategy (fixed-point quantization + differential pair number representation + full precision ADC).

Don’t Need All Eggs in One Basket: Reconstructing Composite Embeddings of Customers from Individual-Domain Embeddings

Although building a 360-degree comprehensive view of a customer has been a long-standing goal in marketing, this challenge has not been successfully addressed in many marketing applications because fractured customer data stored across different “silos” are hard to integrate under “one roof” for several reasons. Instead of integrating customer data, in this article we propose to integrate several domain-specific partial customer views into one consolidated or composite customer profile using a Deep Learning-based method that is theoretically grounded in Kolmogorov’s Mapping Neural Network Existence Theorem. Furthermore, our method needs to securely access domain-specific or siloed customer data only once for building the initial customer embeddings. We conduct extensive studies on two industrial applications to demonstrate that our method effectively reconstructs stable composite customer embeddings that constitute strong approximations of the ground-truth composite embeddings obtained from integrating the siloed raw customer data. Moreover, we show that these data-security preserving reconstructed composite embeddings not only perform as well as the original ground-truth embeddings but significantly outperform partial embeddings and state-of-the-art baselines in recommendation and consumer preference prediction tasks.

Efficient Compilation and Mapping of Fixed Function Combinational Logic onto Digital Signal Processors Targeting Neural Network Inference and Utilizing High-level Synthesis

Recent efforts for improving the performance of neural network (NN) accelerators that meet today’s application requirements have given rise to a new trend of logic-based NN inference relying on fixed function combinational logic. Mapping such large Boolean functions with many input variables and product terms to digital signal processors (DSPs) on Field-programmable gate arrays (FPGAs) needs a novel framework considering the structure and reconfigurability of DSP blocks during this process. The proposed methodology in this article maps the fixed function combinational logic blocks to a set of Boolean functions where Boolean operations corresponding to each function are mapped to DSP devices rather than look-up tables on the FPGAs to take advantage of the high performance, low latency, and parallelism of DSP blocks. This article also presents an innovative design and optimization methodology for compilation and mapping of NNs, utilizing fixed function combinational logic to DSPs on FPGAs employing high-level synthesis flow. Our experimental evaluations across several datasets and selected NNs demonstrate the comparable performance of our framework in terms of the inference latency and output accuracy compared to prior art FPGA-based NN accelerators employing DSPs.

Functional connectivity alterations in traumatic brain injury patients with late seizures

PTE is a neurological disorder characterized by recurrent and spontaneous epileptic seizures. PTE is a major public health problem occurring in 2–50% of TBI patients. Identifying PTE biomarkers is crucial for the development of effective treatments. Functional neuroimaging studies in patients with epilepsy and in epileptic rodents have observed that abnormal functional brain activity plays a role in the development of epilepsy. Network representations of complex systems ease quantitative analysis of heterogeneous interactions within a unified mathematical framework. In this work, graph theory was used to study resting state functional magnetic resonance imaging (rs-fMRI) and reveal functional connectivity abnormalities that are associated with seizure development in traumatic brain injury (TBI) patients. We examined rs-fMRI of 75 TBI patients from Epilepsy Bioinformatics Study for Antiepileptogenic Therapy (EpiBioS4Rx) which aims to identify validated Post-traumatic epilepsy (PTE) biomarkers and antiepileptogenic therapies using multimodal and longitudinal data acquired from 14 international sites. The dataset includes 28 subjects who had at least one late seizure after TBI and 47 subjects who had no seizures within 2 years post-injury. Each subject’s neural functional network was investigated by computing the correlation between the low frequency time series of 116 regions of interest (ROIs). Each subject’s functional organization was represented as a network consisting of nodes, brain regions, and edges that show the relationship between the nodes. Then, several graph measures concerning the integration and the segregation of the functional brain networks were extracted in order to highlight changes in functional connectivity between the two TBI groups. Results showed that the late seizure-affected group had a compromised balance between integration and segregation and presents functional networks that are hyperconnected, hyperintegrated but at the same time hyposegregated compared with seizure-free patients. Moreover, TBI subjects who developed late seizures had more low betweenness hubs.

Combining long-term circuit mapping and network transcriptomics with SiR-N2c.

An exciting frontier in circuit neuroscience lies at the intersection between neural network mapping and single-cell genomics. Monosynaptic rabies viruses provide a promising platform for the merger of circuit mapping methods with -omics approaches. However, three key limitations have hindered the extraction of physiologically meaningful gene expression profiles from rabies-mapped circuits: inherent viral cytotoxicity, high viral immunogenicity and virus-induced alteration of cellular transcriptional regulation. These factors alter the transcriptional and translational profiles of infected neurons and their neighboring cells. To overcome these limitations we applied a self-inactivating genomic modification to the less immunogenic rabies strain, CVS-N2c, to generate a self-inactivating CVS-N2c rabies virus (SiR-N2c). SiR-N2c not only eliminates undesired cytotoxic effects but also substantially reduces gene expression alterations in infected neurons and dampens the recruitment of innate and acquired immune responses, thus enabling open-ended interventions on neural networks and their genetic characterization using single-cell genomic approaches.

History and Future Directions of rTMS for Treatment of Depressive Disorders

History and Future Directions of rTMS for Treatment of Depressive Disorders

Imputation of Gold Recovery Data from Low Grade Gold Ore Using Artificial Neural Network

In a multivariate database, the missing data can be obtained through several imputation techniques, which are particularly useful for data that are difficult to obtain, for any reason, or have high uncertainties or scarce variables. A Self-Organizing Maps (SOM) neural network is an effective tool for the analysis of multidimensional data applied for the imputation of data. In this paper, data from drilling were used for training, testing, and validation using the variables: total Au recovery (%), which means gold recovery from a gravity concentration plus hydrometallurgical process, Au (g/t), As (ppm), S (%), Al2O3 (%), CaO (%), K2O (%), and MgO (%). After training, the partial omission of Au content and recovery was carried out, from 10% to 50%, to evaluate the data imputation performance for those variables. The results imputed by the SOM were compared with the original data values and evaluated according to descriptive statistics; the results indicated a determination coefficient of 85% when 50% of the data were omitted and 93% when 10% of the data were omitted. Once demonstrated, the correlation between the original data and SOM imputation analysis can help geologists and metallurgists to obtain results with a high degree of reliability of metallurgical recovery through related chemical variables, making it possible to implement SOM analysis as a powerful tool to input analytical data. One of the practical applications of the proposed model is to produce a pattern of imputed data that can be a good alternative in the construction or generation of a synthetic geometallurgical database with missing data.