Sequential Learning Algorithm Research Articles

Performance Comparison of Different HTM-Spatial Pooler Algorithms Based on Information-Theoretic Measures

Hierarchical temporal memory (HTM) is a promising unsupervised machine-learning algorithm that models key principles of neocortical computation. One of the main components of HTM is the spatial pooler (SP), which encodes binary input streams into sparse distributed representations (SDRs). In this paper, we propose an information-theoretic framework for the performance comparison of HTM-spatial pooler (SP) algorithms, specifically, for quantifying the similarities and differences between sparse distributed representations in SP algorithms. We evaluate SP's standalone performance, as well as HTM's overall performance. Our comparison of various SP algorithms using Renyi mutual information, Renyi divergence, and Henze–Penrose divergence measures reveals that the SP algorithm with learning and a logarithmic boosting function yields the most effective and useful data representation. Moreover, the most effective SP algorithm leads to superior HTM results. In addition, we utilize our proposed framework to compare HTM with other state-of-the-art sequential learning algorithms. We illustrate that HTM exhibits superior adaptability to pattern changes over time than long short term memory (LSTM), gated recurrent unit (GRU) and online sequential extreme learning machine (OS-ELM) algorithms. This superiority is evident from the lower Renyi divergence of HTM (0.23) compared to LSTM6000 (0.33), LSTM3000 (0.38), GRU (0.41), and OS-ELM (0.49). HTM also achieved the highest Renyi mutual information value of 0.79, outperforming LSTM6000 (0.73), LSTM3000 (0.71), GRU (0.68), and OS-ELM (0.62). These findings not only confirm the numerous advantages of HTM over other sequential learning algorithm, but also demonstrate the effectiveness of our proposed information-theoretic approach as a powerful framework for comparing and evaluating various learning algorithms.

Non-line-of-sight target tracking with improved recurrent extreme learning machine

Target tracking provides important location-based services in many applications. The main challenge of target tracking is to combat the severe degradation problem in Non-Line-of-Sight (NLOS) scenario. Most Deep Learning algorithms available in literature to address this issue belong to batch learning with high complexity. This paper proposes a novel online sequential learning algorithm, Improved Recurrent Extreme Learning Machine (IRELM), to solve the NLOS target tracking problem as a position series prediction task. IRELM is able to train Recurrent Neural Network (RNN) inputs one-by-one and adaptively update the input weight, hidden weight, feedback weight and output weight. Extensive simulations and experiments prove the superior tracking performance and feasible complexity of IRELM over the state-of-the-art Deep Learning methods.

Hyperbolic function based hybrid consistent adaptive sequential DL algorithms for PIR sensor based intruder detection

This study demonstrates that a novel dictionary learning (DL) algorithm-based approach can be successfully used in conjunction with an indigenously developed hardware module, using four pyroelectric infrared (PIR) sensors, to enhance the performance of an intruder detection system. In this work, initially, a hyperbolic function based Consistent Adaptive Sequential Dictionary Learning (CAS-DL) algorithm has been proposed (HCAS-DL). Then another dictionary learning algorithm, called label consistency based K-SVD (LCK-SVD) is considered, where the label information of each dictionary atom has been introduced to enforce discriminability in sparse code. The objective function in LCK-SVD is formuated simultaneously considering the concept of label consistency (LC) constraints, reconstruction errors, and classification errors. Next, the newly proposed HCAS-DL is hybridized with LCK-SVD to form a novel version of hybrid consistent adaptive sequential dictionary learning approach, named here as the LC-HCAS-DL algorithm. Extensive experiments have been performed to establish the suitability of our proposed approach for the problem under consideration. The performance evaluation clearly establishes that the LC-HCAS-DL algorithm can achieve superior performance compared to recently proposed LC-MCAS-DL algorithm and other state-of-the-art algorithms for the intruder detection problem under consideration.

FPGA implementation of two multilayer perceptron neural network in cascade for efficient real time hand gestures tracking

This paper presents implementation of a hand gestures tracking system with a fully connected multilayer perceptron and a supervised sequential learning algorithm on low cost FPGAs (Field Programmable Gate Array). Two networks are used in the study. The first network is trained to detect skins and the second identifies face and hand gestures. Also, a binary image is made from the first network, which is fed into the second network as input. Skin network has two output classes, and hand network 6. The first network recognizes skin and non-skin images and the other distinguishes hands from faces while hand gestures are tracked and classified. Both networks are trained in Matlab before each was implemented on a separate FPGA and the two FPGAs are then connected to each other. Finally, the overall result is compared. Both networks are implemented on Xilinx Spartan XC3S1000 4fg456. Two hardware-implemented networks are applied to a video sequence frame by frame. Features of both hands are extracted and the network is trained using a constructed vector to track left and right hands. Results show that the implemented system enjoys a suitable trade-off between accuracy and speed, reaching 208 fps with the least memory usage. This method has been compared to other methods, as well as methods in which only one threshold is used to detect skin color.

Robust kernel adaptive filtering for nonlinear time series prediction

Recently, online learning algorithms in machine learning have been imposed much attention. As a typical family, kernel adaptive filtering algorithms receive particular interest due to their sequential learning-based features. However, the kernel least mean square (KLMS) algorithm is not suitable for nonlinear tasks corrupted by non-Gaussian noise, especially impulsive noise. This is because the derivation of the KLMS algorithm is on the basis of the mean square error (MSE) criterion which only captures information of second-order statistics. In this paper, motivated by tanh function, we develop its generalized variant by introducing a scale factor for better representation capability; then we incorporate kernel adaptive filter with the generalized tanh function to propose a robust sequential learning algorithm. Based on establishing the energy conservation relation, we derive a sufficient condition for ensuring the algorithm convergence. In addition, to perform the steady-state excess mean square error (EMSE) analysis, we use the pre-tuned dictionary strategy to model the unknown nonlinear system in form of a finite-order combination; by Taylor expansion, we arrive at a closed-form solution for predicting the steady-state behavior. To further improve the algorithm performance, we design an optimization scheme for scale factor. Simulations for nonlinear time series prediction show that the designed schemes yield better performance than some state-of-art algorithms. The steady-state EMSE analysis is validated to provide accurate prediction results.

The Feasibility and Performance of Total Hip Replacement Prediction Deep Learning Algorithm with Real World Data.

(1) Background: Hip degenerative disorder is a common geriatric disease is the main causes to lead to total hip replacement (THR). The surgical timing of THR is crucial for post-operative recovery. Deep learning (DL) algorithms can be used to detect anomalies in medical images and predict the need for THR. The real world data (RWD) were used to validate the artificial intelligence and DL algorithm in medicine but there was no previous study to prove its function in THR prediction. (2) Methods: We designed a sequential two-stage hip replacement prediction deep learning algorithm to identify the possibility of THR in three months of hip joints by plain pelvic radiography (PXR). We also collected RWD to validate the performance of this algorithm. (3) Results: The RWD totally included 3766 PXRs from 2018 to 2019. The overall accuracy of the algorithm was 0.9633; sensitivity was 0.9450; specificity was 1.000 and the precision was 1.000. The negative predictive value was 0.9009, the false negative rate was 0.0550, and the F1 score was 0.9717. The area under curve was 0.972 with 95% confidence interval from 0.953 to 0.987. (4) Conclusions: In summary, this DL algorithm can provide an accurate and reliable method for detecting hip degeneration and predicting the need for further THR. RWD offered an alternative support of the algorithm and validated its function to save time and cost.

Dual stream network with attention mechanism for photovoltaic power forecasting

The operations of renewable power generation systems highly depend on precise Photovoltaic (PV) power forecasting, providing significant economic, and environmental advantages for energy efficient buildings and urban energy systems. However, precise PV power forecasting, particularly, solar power is more challenging due to solar energy intermittence, instability, and randomness. These challenges hinder the integration of PV into smart grids, where accurate power forecasting is a promising solution in this direction, providing effective planning and management services. Therefore, in this work, we introduce a dual-stream network for accurate PV forecasting. The proposed network parallelly learns spatial patterns using convolutional network and temporal representations via sequential learning algorithm. These features are then integrated together to form a single, yet representative feature vector used as an input to self-attention mechanism to further select the optimal features for PV power forecasting. To the best of our knowledge, the proposed dual stream network with advanced features selection mechanism is a pioneering approach for time series analysis, narrowed towards PV power forecasting. We derive our network after a series of experimentations involving solo and hybrid models, resulting in higher forecasting accuracy against state-of-the-art models.

A sequential deep learning algorithm for sampled mixed-integer optimisation problems

Mixed-integer optimisation problems can be computationally challenging. Here, we introduce and analyse two efficient algorithms with a specific sequential design that are aimed at dealing with sampled problems within this class. At each iteration step of both algorithms, we first test the feasibility of a given test solution for each and every constraint associated with the sampled optimisation at hand, while also identifying those constraints that are violated. Subsequently, an optimisation problem is constructed with a constraint set consisting of the current basis—namely, the smallest set of constraints that fully specifies the current test solution—as well as constraints related to a limited number of the identified violating samples. We show that both algorithms exhibit finite-time convergence towards the optimal solution. Algorithm 2 features a neural network classifier that notably improves the computational performance compared to Algorithm 1. We quantitatively establish these algorithms' efficacy through three numerical tests: robust optimal power flow, robust unit commitment, and robust random mixed-integer linear program.

Using biospeckle and LIBS techniques with artificial intelligence to monitor phthalocyanine-gold nanoconjugates as a new drug delivery mediator for in vivo PDT

Photodynamic therapy (PDT) is a promising treatment method that generates reactive oxygen species by exciting a photosensitizer in the presence of light. The positive treatment outcome of PDT is enhanced by the photosensitizer’s ability to get out quickly from normal tissue. This study non-destructively evaluated the biocompatibility of novel synthesized phthalocyanine-gold nanoconjugates (Pc-Au NCs) as a drug delivery mediator for in vivo PDT. We first proposed a novel noncontact and safe biospeckle optical imaging technique to visualize the quantitative changes in the internal dynamics of biological tissue in response to PDT treatment. Second, laser-induced breakdown spectroscopy (LIBS) was utilized to emphasize the accumulation of gold nanoparticles (Au NPs) within the normal and cancerous tissue, which in turn verifies the cytotoxic and treatment effects of the Pc-Au NCs. The biospeckle images and LIBS spectra were obtained from tumor and normal tissues of male BALB/c malignancy-bearing mice before and after exposure to PDT. Then, they were statistically treated via different artificial intelligence techniques. For biospeckle analysis, the proposed method was established from the combination of second and higher-order statistical features extracted from the gray-level co-occurrence matrix (GLCM) and local contrast analysis (LCA), respectively, through principal component analysis (PCA). Regarding LIBS analysis, the powerful, most popular ensemble sequential learning algorithm Adaptive Boost was implemented on highly uncorrelated variables extracted from the raw spectra. The results, confirmed by the histological investigation, indicated the potential of the biospeckle and LIBS statistical properties in demonstrating the efficiency of the Pc-Au NCs as a drug delivery mediator for in vivo PDT.

Development of an electronic frailty index for predicting mortality in patients undergoing transcatheter aortic valve replacement using machine learning

Background: Electronic frailty indices can be useful surrogate measures of frailty. Objective: This study is to develop an electronic frailty index that incorporates patient demographics, baseline comorbidities, health-care utilization characteristics, electrocardiographic measurements, and laboratory examinations for predicting all-cause mortality in patients undergoing transcatheter aortic valve replacement (TAVR). Methods: This was a multicenter retrospective observational study of patients undergoing for TAVR. Significant univariate and multivariate predictors of all-cause mortality were identified using Cox regression. Importance ranking of variables was obtained with a gradient boosting survival tree (GBST) model, a supervised sequential ensemble learning algorithm, and used to build the frailty models. Comparisons were made between multivariate Cox, GBST, and random survival forest models. Results: A total of 450 patients (49% of females; median age at procedure, 82.3 [interquartile range, 79.0–86.0]) were included, of which 22 died during follow-up. A machine learning survival analysis model found that the most important predictors of mortality were activated partial thromboplastin time, followed by INR, severity of tricuspid regurgitation, cumulative hospital stays, cumulative number of readmissions, creatinine, urate, alkaline phosphatase, and QTc/QT intervals. GBST significantly outperformed random survival forests and multivariate Cox regression (precision: 0.91, recall: 0.89, AUC: 0.93, C-index: 0.96, and KS-index: 0.50) for mortality prediction. Conclusions: An electronic frailty index incorporating multidomain data can efficiently predict all-cause mortality in patients undergoing TAVR. A machine learning survival learning model significantly improves the risk prediction performance of the frailty models.

Multi-level classification technique for diagnosing osteoporosis and osteopenia using sequential deep learning algorithm

Multi-level classification technique for diagnosing osteoporosis and osteopenia using sequential deep learning algorithm

Leverage effect in cryptocurrency markets

In this article we study the leverage effect in cryptocurrency markets using a stochastic volatility model with simultaneous and correlated jumps in returns and volatility. We estimate the model using an efficient sequential learning algorithm with daily data on four actively traded cryptocurrencies including Bitcoin, Ethereum, Chainlink, and Litecoin. Doing so allows us to sequentially learn about the return-volatility relationships and the leverage effect in these cryptocurrencies when new data come in. We find that these relationships depend on both the diffusive and jump components of correlations between returns and volatility. Interestingly, the diffusive and jump components often have opposite signs for these currencies; that is, while the diffusive component may exhibit a negative return-volatility relationship (the “leverage effect”), the jump component may show a positive relationship (the “inverse leverage effect”). As a result, the total leverage effect can be quite different from the diffusive leverage effect, due to the presence of correlated jumps in returns and volatility. Overall, we provide evidence that these jumps matter greatly to the total leverage effect in cryptocurrency markets.

Classification Method of Voltage Sag Sources Based on Sequential Trajectory Feature Learning Algorithm

The classification of voltage sag sources is essential for the establishment of controlling scheme and reasonable division of responsibilities in voltage sag-associated accidents. Existing methods for classifying voltage sag sources usually ignore the interpretability of the classification model, and are only dedicated to improving the accuracy of voltage classification, which cannot provide a reliable classification basis for users and power enterprises. Therefore, this article proposes an effective and interpretable voltage sag sources classification method based on sequential trajectory feature learning and Random Forest algorithm. Firstly, to fully consider the interpretable sequential trajectory features of voltage sag signals so as to improve the quality and interpretability of calculation process, the fused lasso generalized eigenvector (FLAG) algorithm is adopted to quickly search for interpretable shapelets sub-sequences from the labeled voltage sag data. After that, the labeled data and samples to-be-classified are subjected to shapelet transformation through the shapelet sub-sequences to obtain the sequential trajectory features. Finally, the random forest is trained on the labeled sequential trajectory feature data to achieve supervised sample classification, which inherits the interpretability of shapelet. To test the feasibility and validity of the proposed voltage sag sources classification method, the simulation cases based on the simulated voltage sag signals were studied. The simulation results show that the proposed method has significant advantages in terms of accuracy and interpretability of voltage sag sources classification.

Derivation of an electronic frailty index for short-term mortality in heart failure: a machine learning approach

Abstract Objective Frailty may be found in heart failure patients especially in the elderly and is associated with a poor prognosis. However, assessment of frailty status is time-consuming and the electronic frailty indices developed using health records have served as useful surrogates. We hypothesized that an electronic frailty index developed using machine learning can improve short-term mortality prediction in patients with heart failure. Methods This was a retrospective observational study included patients admitted to nine public hospitals for heart failure from Hong Kong between 2013 and 2017. Age, sex, variables in the modified frailty index, Deyo's Charlson comorbidity index (≥2), neutrophil-to-lymphocyte ratio (NLR) and prognostic nutritional index (PNI) were analyzed. Gradient boosting, which is a supervised sequential ensemble learning algorithm with weak prediction submodels (typically decision trees), was applied to predict mortality. Variables were ranked in the order of importance with a total score of 100 and used to build the predictive models. Comparisons were made with decision tree and multivariate logistic regression. Results A total of 8893 patients (median: age 81, Q1-Q3: 71–87 years old) were included, in whom 9% had 30-day mortality and 17% had 90-day mortality. PNI, age and NLR were the most important variables predicting 30-day mortality (importance score: 37.4, 32.1, 20.5, respectively) and 90-day mortality (importance score: 35.3, 36.3, 14.6, respectively). Gradient boosting significantly outperformed decision tree and multivariate logistic regression (area under the curve: 0.90, 0.86 and 0.86 for 30-day mortality; 0.92, 0.89 and 0.86 for 90-day mortality). Conclusions The electronic frailty index based on comorbidities, inflammation and nutrition information can readily predict mortality outcomes. Their predictive performances were significantly improved by gradient boosting techniques. Funding Acknowledgement Type of funding sources: None.

Derivation of an electronic frailty index for predicting short‐term mortality in heart failure: a machine learning approach

AimsFrailty may be found in heart failure patients especially in the elderly and is associated with a poor prognosis. However, assessment of frailty status is time‐consuming, and the electronic frailty indices developed using health records have served as useful surrogates. We hypothesized that an electronic frailty index developed using machine learning can improve short‐term mortality prediction in patients with heart failure.Methods and resultsThis was a retrospective observational study that included patients admitted to nine public hospitals for heart failure from Hong Kong between 2013 and 2017. Age, sex, variables in the modified frailty index, Deyo's Charlson co‐morbidity index (2), neutrophil‐to‐lymphocyte ratio (NLR), and prognostic nutritional index at baseline were analysed. Gradient boosting, which is a supervised sequential ensemble learning algorithm with weak prediction submodels (typically decision trees), was applied to predict mortality. Variables were ranked in the order of importance with a total score of 100 and used to build the frailty models. Comparisons were made with decision tree and multivariable logistic regression. A total of 8893 patients (median: age 81, Q1–Q3: 71–87 years old) were included, in whom 9% had 30 day mortality and 17% had 90 day mortality. Prognostic nutritional index, age, and NLR were the most important variables predicting 30 day mortality (importance score: 37.4, 32.1, and 20.5, respectively) and 90 day mortality (importance score: 35.3, 36.3, and 14.6, respectively). Gradient boosting significantly outperformed decision tree and multivariable logistic regression. The area under the curve from a five‐fold cross validation was 0.90 for gradient boosting and 0.87 and 0.86 for decision tree and logistic regression in predicting 30 day mortality. For the prediction of 90 day mortality, the area under the curve was 0.92, 0.89, and 0.86 for gradient boosting, decision tree, and logistic regression, respectively.ConclusionsThe electronic frailty index based on co‐morbidities, inflammation, and nutrition information can readily predict mortality outcomes. Their predictive performances were significantly improved by gradient boosting techniques.

Simulation And Modelling of BCI Based Multi Purpose Wheel Chair For Paralysed People

According to the survey of Reeve foundation and WHO says, there are nearly 1 in 50 people living with paralysis, every year 5.4 million people affected. Paralysis is caused by spinal injury, stroke, multiple sclerosis, cerebral palsy and other causes like motor accidents and victim of violence. It is also called loss of muscle function in some part of the body. In this paper, we systematically survey a different techniques used for brain computer interface and also review the research on non-invasive, electroencephalography (EEG)-based BCI systems for communication and rehabilitation. Main research focus on previous techniques like deep learning or deep neural network, machine learning, Neuro plasticity, support vector machine, artifact suppression and so on. In this Differing from traditional machine learning algorithms, online sequential machine learning algorithm is empowered to learn distinct high-level representations from raw brain signals without manual feature selection. A certain MATLAB program is designed to use these motions. Our results are to exploit the accuracy rate and to generate the assistive devices for restoration of movement and communication strength for physically disabled patients in order to rehabilitate their lost motor abilities.

Electronic frailty index for predicting mortality outcome of patients undergoing transaortic valvular implantation

Abstract Funding Acknowledgements Type of funding sources: None. Objective Electronic frailty index for predicting mortality outcome of patients undergoing transaortic valvular implantation (TAVI) served as useful surrogates but is associated with a poor prognosis since it needs long time to determine the frailty status and develop the index based on electronic health records. We identify significant risk mortality predictors and tested the hypothesis that an electronic frailty index incorporating ECG measurements and laboratory examinations using a machine learning survival analysis approach can improve TAVI mortality prediction. Design A territory-wide observational study which involved a total of 450 patients (49.11% females, 22 mortalities) diagnosed undergoing TAVI and admitted to public hospitals from Hong Kong. Methods Demographics (TAVI presentation age, gender, severity of TR, AR, MR, PR, INR of TAVI presentation), prior comorbidities before TAVI presentation, ECG measurements, and CBC and LRFT laboratory examinations were analyzed. Cox regression and a supervised sequential ensemble learning algorithm: gradient boosting survival tree (GBST) model, was applied to predict mortality. Significant univariate and multivariate risk predictors of mortality were identified. Importance ranking of variables were obtained with GBST model and used to build the frailty models. Comparisons were provided with baseline models of random survival forests and multivariate Cox regression. Results The median TAVI presentation age was 82.3 years (83.8 years in mortalities, and 82.1 years in alive patients). INR of TAVI presentation in mortalities (median: 1.32) is much higher than alive ones (median: 1.07). Severe TR (hazard ratio, HR: 8.93, 95% CI: [3.22, 24.78], p value < 0.0001), INR of TAVI presentation (HR: 2.74, 95% CI: [1.84, 4.09], p value < 0.0001), cumulative hospital stays (HR: 1.01, 95% CI: [1.00, 1.01], p value = 0.0008), aspartate transaminase (HR: 1.01, 95% CI: [0.98, 1.002], p value = 0.0002), and bilirubin (HR: 1.02, 95% CI: [1.01, 1.02], p value = 0.0003) are significant mortality risk predictors. Machine learning survival analysis model found that APTT demonstrates the most important strength, followed by INR of TAVI presentation, severe TR status, cumulative hospital stays, cumulative readmission times, creatinine test, urate test ALP test, and ECG measurements of QTc and QT. GBST significantly outperformed random survival forests and multivariate Cox regression (precision: 0.91, recall: 0.89, AUC: 0.93, C-index: 0.96, and KS-index: 0.50) for mortality prediction. Conclusions Electronic frailty index based on demographics, prior comorbidities, hospitalization characteristics, ECG measurements, and laboratory examinations can efficiently predict mortality outcome of patients undergoing TAVI. Machine learning survival learning model significantly improves the risk prediction performance and improves the construction of the frailty models for tailored interventions of TAVI patients in clinical practices.

In data science we trust: Machine learning for stable halide perovskites

The exploration of the compositional space of halide perovskites is prohibitively costly using traditional experimental research strategies. Sun and colleagues demonstrated a physics-informed, machine-learning-guided, experimental approach that overturns this limitation enabling a rapid and efficient maximization of stability in mixed-cation perovskites.

Sequential Learning of Cryptocurrency Volatility Dynamics: Evidence Based on a Stochastic Volatility Model with Jumps in Returns and Volatility

This paper studies the dynamics of cryptocurrency volatility using a stochastic volatility model with simultaneous and correlated jumps in returns and volatility. We estimate the model using an efficient sequential learning algorithm that allows for learning about multiple unknown model parameters simultaneously, with daily data on four popular cryptocurrencies. We find that these cryptocurrencies have quite different volatility dynamics. In particular, they exhibit different return-volatility relationships: While Ethereum and Litecoin show a negative relationship, Chainlink displays a positive one and interestingly, Bitcoin’s one changes from negative to positive in June 2016. We also provide evidence that the sequential learning algorithm helps better detect large jumps in the cryptocurrency market in real time. Overall, incorporating volatility jumps helps better capture the dynamic behavior of highly volatile cryptocurrencies.

A novel online sequential extreme learning machine with L2,1-norm regularization for prediction problems

In today’s world, data is produced at a very high speed and used in a large number of prediction problems. Therefore, the sequential nature of learning algorithms is in demand for batch learning algorithms. This paper presents a novel online sequential algorithm for extreme learning machine with l2,1-norm regularization (LR21OS-ELM) to handle the real-time sequential data. Wang et al. have given ELM with l2,1-norm based regularization namely LR21-ELM. This method is a batch processing model which takes data in a single chunk. So, whenever a new chunk of data arrives the model has to be retrained which takes a lot of time and memory. The proposed sequential algorithm does not require building a new model each time data arrives. This will update the previous model with new data that will save time and memory. The l2,1-norm regularization is a structural sparse-inducing norm which is integrated with an online sequential learning algorithm to diminish the complexity of the learning model by eliminating the redundant neurons of OS-ELM model. This paper proposes an iterative bi-objective optimization algorithm to solve l2,1 norm-based minimization problem and to handle the real time sequential data. The proposed model can learn sequentially arriving data in the form of chunks where chunk size can be fixed or varying. The experimental study has been conducted on several benchmark datasets collected from different research domains to prove the generalization ability of the proposed algorithm. The obtained results show that LR21OS-ELM combines the advantages of l2,1-norm regularization and online sequential learning of data and improves the prediction performance of the system.