Inference System Research Articles

Nitrogen‐Doped Graphene Nanosheets for Sustainable Removal of Pharmaceutical Contaminants from Water: Kinetics and ANFIS Modeling

AbstractIn this work, nitrogen‐doped graphene nanosheets (NGS) were synthesized to remove diclofenac sodium (DFS) from water. The NGS exhibited 99.8 % removal efficiency, adsorption capacity of 278 mg g−1 for DFS within 30 min, and excellent reusability up to 5 cycles of adsorption. The experimental data fit well with pseudo‐second order kinetic model and Langmuir isotherm model. X‐ray photoelectron spectroscopy (XPS) and FTIR studies revealed that the carboxylic group, triazine group, graphitic‐N, and pyridinic‐N groups were involved in the adsorption of DFS. The adaptive neuro‐fuzzy inference system (ANFIS) was modeled for accurate prediction of optimizing parameters for the maximum adsorptive removal (%) of DFS by NGS. A subsequent cytotoxicity assay of NGS on a green alga Chlamydomonas sp. signified non‐toxic nature. This study suggests NGS to be a promising economic adsorbent for water remediation in resource‐poor environments to fulfill “UN‐Sustainable Development Goal‐ 6”.

Modelling using fuzzy logic: study of needle-punched filters

The tensile behaviour dependency on the process parameters of blended nonwoven fabric is highly non-linear. This paper explores the potentiality of fuzzy logic inference in such a nonlinear environment as predictive process modelling with minimal experimental data. A model based on fuzzy inference engine has been attempted using recycled polyester fibre blend percentage, needling density and depth of needle penetration as the input variables and tensile strength in both machine direction (MD) and cross direction (CD) as the output variables. Proposed model was built in MAT LAB/Simulink and has been verified by experimental sample data set. Mean relative error percentage and correlation coefficient between the fuzzy model predicted and actual experimental sample values were found to be around ±2.5% and up to 0.97 respectively. These results validate that the model may be applied with high accuracies for the prediction of tensile behaviour of the blended nonwoven fabric in the textile industries.

An interpretable model for stock price movement prediction based on the hierarchical belief rule base

Stock price movement prediction is the basis for decision-making to maintain the stability and security of stock markets. It is important to generate predictions in an interpretable manner. The Belief Rule Base (BRB) has certain interpretability based on IF-THEN rule semantics. However, the interpretability of BRB in the whole process of stock prediction modeling may be weakened or lost. Therefore, this paper proposes an interpretable model for stock price movement prediction based on the hierarchical Belief Rule Base (HBRB-I). The interpretability of the model is considered, and several criteria are constructed based on the BRB expert system. First, the hierarchical structure of BRB is constructed to ensure the interpretability of the initial modeling. Second, the interpretability of the inference process is ensured by the Evidential Reasoning (ER) method as a transparent inference engine. Third, a new Projection Covariance Matrix Adaptive Evolution Strategy (P-CMA-ES) algorithm with interpretability criteria is designed to ensure the interpretability of the optimization process. The final mean squared error value of 1.69E-04 was obtained with similar accuracy to the initial BRB and enhanced in terms of interpretability. This paper is for short-term stock forecasting, and more data will be collected in the future to update the rules to enhance the forecasting capability of the rule base.

Convergent Dynamics In Physiological Response As A Window-Based Robust Control System

The process adapted here delivers physiological readings and maps out the major components that make up the physiological system. The contribution of this component can also be determined on each of the parameters. The skin conductance response (SCR) has a major effect on the synchronous readings of the physiological metrics used. The paper investigates this default by developing a prototypical representation of integrated physiological readings with real-time analytics using a convergent dynamic control model as the inference engine. The major components that make up the physiological system and errors, are characterised by either their noisiness or systematic disturbance. The noisy error is based on the obvious pattern in each time interval and it is measured based on the mean root square error(MRSE). The SCR has a major effect on the synchronous readings of the physiological metrics used. The results obtained from test runs indicate that the Tonic phase of the physiological response signal is the skin conductance level (SCL) that represents the base level of the signal. The Phasic level is the component of the skin conductance response that reflects the direct response to the external stimulus, and this is set between 1-4secs after the stimulus onset at the baseline phase. There also exists the non-specific SCR known as the anomalies that appear post-stimuli; this represents the number of conductance responses that appear within and among subsequent physiological response readings. The originality of work is its ability to visualise multi-modal response signals in a single frame and detect optimal responses significant to a particular event-related protocol

A visual analytics workflow for probabilistic modeling

Probabilistic programming is a powerful means for formally specifying machine learning models. The inference engine of a probabilistic programming environment can be used for serving complex queries on these models. Most of the current research in probabilistic programming is dedicated to the design and implementation of highly efficient inference engines. Much less research aims at making the power of these inference engines accessible to non-expert users. Probabilistic programming means writing code. Yet many potential users from promising application areas such as the social sciences lack programming skills. This prompted recent efforts in synthesizing probabilistic programs directly from data. However, working with synthesized programs still requires the user to read, understand, and write some code, for instance, when invoking the inference engine for answering queries. Here, we present an interactive visual approach to synthesizing and querying probabilistic programs that does not require the user to read or write code.

EXPERT SYSTEM FOR DIAGNOSING DISEASE USING FUZZY LOGIC

In this research, a system for diagnosing disease was designed and implemented. This system will be used for paramedics expert. They need to enter many symptoms. The symptoms are classified into two categories, crisps and fuzzy. The fuzzy symptoms are input using slider. It is used for uncertainty symptoms. Building this system, needs three steps knowledge acquisition, knowledge representation and then building knowledge base, and finally building inference engine. This system uses standard class operator, algebraic class operator, and Einstein class operator. Implication rule uses Dienes-Rescher and for combine all rules use Mamdani Combination. Based on all symptom that have been entered using standard class operator, this system is a success for diagnosing disease and its inference value. Keywords : diagnose disease, expert system, fuzzy logic.

Significance measures for rules in probabilistic-fuzzy inference systems based on fuzzy transforms

This paper defines a novel probabilistic-fuzzy inference system that considers fuzzy inputs and returns, as output, a probability distribution. In this way, it combines two different ways to represent uncertainty: the one modeled by fuzzy theory, that allows to represent reliable but vague information; and the one modeled by probability theory, that allows to represent undetermined but specific information. The novelty of this probabilistic-fuzzy inference system, with respect to the other existing in the literature, is that its inference engine combines quantile functions instead of distribution, probabilistic or density functions. Besides the formal definition of this novel kind of fuzzy inference systems, we propose: firstly, the construction of probabilistic-fuzzy rules by means of direct quantiles F-transforms; secondly, the definition of several significance measures for the obtained association rules; and finally, we present a set of experiments to validate all the assertions done throughout the paper.

Process Chain-Oriented Design Evaluation of Multi-Material Components by Knowledge-Based Engineering

The design of components suitable for manufacturing requires the application of knowledge about the manufacturing process chain with which the component is to be manufactured. This article presents an assistance system for decision support in the context of design for manufacturing. The assistance system includes explicit manufacturing process chain knowledge and has an inference engine that can automatically evaluate the manufacturability of a component design based on a given manufacturing process chain and resolve emerging manufacturing conflicts by making adjustments on the component or resource side. A link with a CAD system additionally enables the three-dimensional representation of derived manufacturing stages and manufacturing resources. Within the assistance system, a manufacturing process chain is understood as a configurable design object and is implemented via a constraint satisfaction problem. Furthermore, the required abstraction of manufacturing processes within finite domains can be reduced to the extent that necessary modeling resolution is achieved by incorporating empirical or simulative surrogate models into the CSP. The assistance system was conceptually validated on a tailored forming process chain for the production of a multimaterial shaft and provides added value, as valuable manufacturing information for component designs is automatically derived and made available in explicit form during the component development.

On the Implementation of Fuzzy Inference Engines on Quantum Computers

Quantum computers can be a revolutionary tool to implement inference engines for fuzzy rule-based systems. In fact, the use of quantum mechanical principles can enable parallel execution of fuzzy rules and allow them to be used efficiently in complex contexts such as distributed and big data environments. This paper introduces the very first quantum-based fuzzy inference engine that is capable of providing exponential acceleration in fuzzy rule execution compared to its classical counterpart, and allows a quantum computer to be programmed by fuzzy linguistic rules. The proposed inference engine was implemented using a quantum algorithm design scheme based on the oracle notion. This scheme allows the modeling of a fuzzy rule-based system as a Boolean function, the oracle, which is able to reconstruct the relationships between the antecedent and consequent parts of fuzzy rules, and can be efficiently computed on a quantum computer. The suitability of the proposed quantum algorithm for use as a fuzzy inference engine was tested in a typical benchmark scenario, such as that provided by inverted pendulum control.

Embedded fuzzy logic for controlling pH and nutrition in hydroponic cultivation

The uptake of plant nutrients causes nutritional imbalances in hydroponic recirculating systems. Generally, the biological parameters of nutrition solution systems are regulated by pH and nutrients that affect plant health; thus, they must be monitored and managed. In order to solve this issue, a fuzzy logic-based automated control system for intelligent hydroponics was developed. This paper describes the implementation of fuzzy logic controllers into a hydroponic monitoring system. This method produces a controlled hydroponic system in hydroponics by first gathering hydroponic plant data in real time. The system is created using the following steps: system design, prototyping, testing, and assessment. The system’s input consists of TDS and pH sensors, which are used to detect the plant’s nutrient content and pH. The system’s output is a 4-channel relay coupled to 4 small pumps for raising the pH level (pH up), reducing the pH level (pH down), and delivering nutrition AB mix. The fuzzy inference engine employs a combination of the Min-Max and Mamdani techniques. Defuzzification was performed using the centroid technique. The embedded fuzzy logic library (eFLL) is an auxiliary library used by the Arduino IDE. The developed mechanism makes appropriate selections about the nutrition solution management system and pH level.

Analysis and Application of the Certainty Factor Method in the Diagnostic Expert System for Broiler Chicken Diseases

An expert system is a system that uses human knowledge recorded in a computer to solve problems that normally require human expertise. Certainty Factor is a clinical parameter value given by MYCIN to indicate the level of trust. The certainty factor is also a way of combining belief and disbelief in a single number. Certainty factor introduces the concepts of belief/belief and disbelief/unbelief. In the system analysis stage, the author analyzes to find weaknesses in the system that is currently running in an organization or company with the aim that the system can be proposed for improvement. Implementing the program that has been designed requires a tool in the form of a computer or laptop, which requires three supporting components to operate the computer or laptop itself, such as hardware, software, and brainware. The expert system application has been running with the Certainty Factor method and to make a diagnosis on chickens the user can choose a value according to the Certainty Factor Interpretation that has been given and processed with the CF expert. Based on the tests carried out, the value of system accuracy with experts is almost the same . And the Inference engine works well, according to the pre-programmed rules. After diagnosing a chicken disease, the system will display how to treat and handle sick broiler chickens

Dóris Chatbot: Personal assistant to help indicate a medical specialty

Conversational interfaces have gained visibility due to recent technological advances in the use of Artificial Intelligence. In the health area, this type of interface can be very useful for applications that directly interact with the patient, therefore, simulating contact with a health agent. In addition, many patients, when having certain symptoms, usually make a partial self-diagnosis, without a clinical basis, and make an appointment with a specialist doctor, without first asking the opinion of a general practitioner. In this context, the objective of the research was to develop a chatbot platform in the health area, using the agent Dóris, with the functionality of pre-medical consultation to indicate the most appropriate medical specialty for patients with cough. The platform was developed using IBM Watson for natural language processing together with an inference engine for decision making.

Comparison of nursing diagnostic accuracy when aided by Knowledge-Based Clinical Decision Support Systems with Clinical Diagnostic Validity and Bayesian Decision Models for psychiatric care plan formulation among nursing students: a quasi-experimental study

BackgroundThe most suitable and reliable inference engines for Clinical Decision Support Systems in nursing clinical practice have rarely been explored.PurposeThis study examined the effect of Clinical Diagnostic Validity-based and Bayesian Decision-based Knowledge-Based Clinical Decision Support Systems on the diagnostic accuracy of nursing students during psychiatric or mental health nursing practicums.MethodsA single-blinded, non-equivalent control group pretest–posttest design was adopted. The participants were 607 nursing students. In the quasi-experimental design, two intervention groups used either a Knowledge-Based Clinical Decision Support System with the Clinical Diagnostic Validity or a Knowledge-Based Clinical Decision Support System with the Bayesian Decision inference engine to complete their practicum tasks. Additionally, a control group used the psychiatric care planning system without guidance indicators to support their decision-making. SPSS, version 20.0 (IBM, Armonk, NY, USA) was used for data analysis. chi-square (χ2) test and one-way analysis of variance (ANOVA) used for categorical and continuous variables, respectively. Analysis of covariance was done to examine the PPV and sensitivity in the three groups.ResultsResults for the positive predictive value and sensitivity variables indicated that decision-making competency was highest in the Clinical Diagnostic Validity group, followed by the Bayesian and control groups. The Clinical Diagnostic Validity and Bayesian Decision groups significantly outperformed the control group in terms of scores on a 3Q model questionnaire and the modified Technology Acceptance Model 3. In terms of perceived usefulness and behavioral intention, the Clinical Diagnostic Validity group had significantly higher 3Q model and modified Technology Acceptance Model 3 scores than the Bayesian Decision group, which had significantly higher scores than the control group.ConclusionKnowledge-Based Clinical Decision Support Systems can be adopted to provide patient-oriented information and assist nursing student in the rapid management of patient information and formulation of patient-centered care plans.

Breast cancer: A hybrid method for feature selection and classification in digital mammography

AbstractIn this article, a hybrid approach based on the Whale optimization algorithm (WOA) and the Dragonfly algorithm (DA) is proposed for breast cancer diagnosis. The hybrid WOADA method selects features based on the fitness value. These features are used to predict the breast masses as benign or malignant using artificial neural networks (ANN) and adaptive neuro‐fuzzy inference systems (ANFIS) as classifiers. The proposed solution is evaluated by using 651 mammograms. The results demonstrate that the WOADA technique outperforms the basic WOA and DA approaches. The accuracy of the suggested WOADA algorithm is 97.84%, with a Kappa value of 0.9477 and an AUC value of 0.972 ± 0.007 for the ANN classifier. Likewise, the ANFIS classifier achieved 98.00% accuracy with a Kappa value of 0.96 and an AUC value of 0.998 ± 0.001. In addition, the viability of the hybrid WOADA technique was evaluated on four benchmark datasets and then compared with four state‐of‐the‐art algorithms and published approaches.

A low cost neuromorphic learning engine based on a high performance supervised SNN learning algorithm

Spiking neural networks (SNNs) are more energy- and resource-efficient than artificial neural networks (ANNs). However, supervised SNN learning is a challenging task due to non-differentiability of spikes and computation of complex terms. Moreover, the design of SNN learning engines is not an easy task due to limited hardware resources and tight energy constraints. In this article, a novel hardware-efficient SNN back-propagation scheme that offers fast convergence is proposed. The learning scheme does not require any complex operation such as error normalization and weight-threshold balancing, and can achieve an accuracy of around 97.5% on MNIST dataset using only 158,800 synapses. The multiplier-less inference engine trained using the proposed hard sigmoid SNN training (HaSiST) scheme can operate at a frequency of 135 MHz and consumes only 1.03 slice registers per synapse, 2.8 slice look-up tables, and can infer about 0.03times {varvec{10}}^{varvec{9}} features in a second, equivalent to 9.44 giga synaptic operations per second (GSOPS). The article also presents a high-speed, cost-efficient SNN training engine that consumes only 2.63 slice registers per synapse, 37.84 slice look-up tables per synapse, and can operate at a maximum computational frequency of around 50 MHz on a Virtex 6 FPGA.

A 218 GOPS neural network accelerator based on a novel cost-efficient surrogate gradient scheme for pattern classification

The accuracy and hardware efficiency of a neural system depends critically on the choice of an activation function. Rectified linear unit (ReLU) is a contemporary activation function that yields high accuracy and allows the construction of efficient neural chips, but it results in a lot of dead neurons, especially at the output layer. This problem is more pronounced in case of multichannel, multiclass classification. This is due to the fact that ReLU cancels out negative values altogether, as a result of which the corresponding values cannot be successfully backpropagated. This phenomenon is referred to as the dying ReLU problem. In this article, we present a novel ‘surrogate gradient’ learning scheme in order to solve gradient vanishing and the dying ReLU problems. To the best of our knowledge, this is the first learning scheme that enables the use of ReLU for all network layers while solving the Dying ReLU problem. We also present a high-performance inference engine that uses ReLU-based actuators for all the network layers in order to achieve high hardware efficiency. The design is excellent for online learning as well, since the derivative of the activation is equal to a constant and can be implemented using low-complexity components. The proposed technique significantly outperforms various contemporary schemes for the CIFAR-10 dataset, and can successfully yield about 98.39% accuracy on MNIST dataset while using less than 159k synapses. Moreover, the proposed hardware implementation is able to perform about 218 giga operation per second (GOPS) while consuming only about 3.95 slice registers and 25.89 slice look-up tables per synapse on a low-end Virtex 6 FPGA. The system is able to operate at a clock frequency of 93.2 MHz.

Comparative evaluation of bioethanol fermentation process parameters using RSM, ANN and ANFIS

AbstractThe drive for renewable energy as an alternative to fossil fuel is on the increase globally. Modeling a renewable energy process is crucial for increasing process monitoring and control of plant efficiency for the ultimate objective of optimal biorefinery operations. This study aims to develop and compare models that best predict the fermentation process parameters of bioethanol production using corn‐steep liquor (CSL) as a media supplement. The response surface method (RSM), artificial neural networks (ANNs) and adaptive neuro‐fuzzy inference system (ANFIS) were tested in the modeling of bioethanol fermentation processes. Box–Behnken design was used to investigate fermentation process parameters considering the effect of CSL (0.5–5.5 w/v), pH (4, 5), time (12–36 h), temperature (25–35°C) and inoculum size (0.5–5.5 v/v). The results from the kinetics study show media formulation with corn steep liquor results in a comparable yield with that substituted with yeast extract. The study shows that, for CSL, a maximum ethanol concentration of 17.40 g L−1 was obtained after 48 h of fermentation while 21.53 g L−1 was attained after 6 h for the media formulation with yeast extract. Based on the model evaluation using statistical error indices, ANN predictability was better at R2 = 0.90; R = 0.95; SEP = 1.73. The ANN models described the process better than ANFIS and RSM. This study shows the intelligent predictive ability of ANN that could be useful for the scale‐up process of ethanol production in industry. © 2023 The Authors. Biofuels, Bioproducts and Biorefining published by Society of Industrial Chemistry and John Wiley & Sons Ltd.

Developing an Inference Engine for Estimating State of Charge of the Lead Acid Battery

Lead acid battery efficiency is affected by huge uncertainty features. An inference engine is required to monitor the uncertainty of the battery state of charge. The objective of the research is to design an inference system to predict the lead acid battery state of charge. A Relief algorithm and Pearson correlation were applied to pre-process the real-world dataset. A fuzzy inference system was adopted to design the inference engine of the state of charge. This research found four main features which had significant impact to lead acid state of charge, including: export power, temperature, volt per cell and ampere. These features had different directions of correlation and furtherly set as inference system’s output. This research had successfully developed an inference engine for lead acid state of charge with Mamdani fuzzy type and centroid defuzzification. In the future, it needs expert validation of the developed rules in the inference system.

Real-time detection of deep-sea hydrothermal plume based on machine vision and deep learning

Recent years have witnessed an increase in applications of artificial intelligence (AI) in the detection of oceanic features with the tremendous success of deep learning. Given the unique biological ecosystems and mineral-rich deposits, the exploration of hydrothermal fields is both scientifically and commercially important. To achieve autonomous and intelligent sampling of the hydrothermal plume by using AUV, this paper proposes an innovative method for real-time plume detection based on the YOLOv5n deep learning algorithm designed with a light-weight neural network architecture to meet the requirements of embedded platforms. Ground truth labeler app LabelImg was used to generate the ground truth data from the plume dataset created by ourselves. To accurately and efficiently detect hydrothermal plumes using an embedded system, we improved the original structure of YOLOv5n in two aspects. First, SiLU activation functions in the model were replaced by ReLU activations at shallow layers and Hard-SiLU activations at deep layers to reduce the number of calculations. Second, an attention module termed Coordinate Attention (CA) was integrated into the model to improve its sensitivity to both channel and spatial features. In addition, a transfer learning training method was adopted to further improve the model’s accuracy and generalizability. Finally, we successfully deployed the proposed model in a low-cost embedded device (NVIDIA Jetson TX2 NX) by using the TensorRT inference engine. We then installed the Jetson TX2 NX into a hovering-type AUV as its vision processing unit and conducted a plume detection test in the water tank. The water tank experimental results demonstrated that the proposed method can achieve real-time onboard hydrothermal plume detection.

Harnessing Deep Neural Networks to Analyze Multi‐Channel Anion Sensing Characteristics of a Ru(II)‐Pyrazolyl‐Bis(Benzimidazole) Complex

AbstractIn this work, the anion‐responsive conduct of a Ru(II)‐bipyridine complex incorporating pyrazolyl‐bis (benzimidazole) ligand is thoroughly investigated in acetonitrile and water via absorption and emission spectroscopy as well as by square‐wave voltammetry (SWV). Substantial alteration of the photo‐redox behavior of the complex is observed in the presence of the selected anions. The free form of the complex exhibits emission indicating the “on‐state”, while inclusion of anions leads to quenching of emission and represents the “off‐state”. The restoration of the initial state of the complex is feasible in the presence of acid and the process is reversible and can be recycled. In essence, the complex functions as anion‐ and acid‐responsive molecular switches. Additionally, we applied herein neural network based deep learning methodologies, viz. Artificial Neural Networks (ANNs) and Adaptive Neuro‐Fuzzy Inference System (ANFIS)} for thorough analysis and fully understand the multi‐channel anion sensing behavior of the complex.