Shallow Learning Models Research Articles

Functional Brain Network Classification for Alzheimer’s Disease Detection with Deep Features and Extreme Learning Machine

The human brain can be inherently modeled as a brain network, where nodes denote billions of neurons and edges denote massive connections between neurons. Analysis on functional brain networks provides powerful abilities to discover potential mechanisms of human brain, and to aid brain disease detection, such as AD (Alzheimer’s disease). Effective discrimination of patients of AD and MCI (mild cognitive impairment) from NC (normal control) is important for the early diagnosis of AD. Therefore, this paper explores the problem of brain network classification for AD detection. Two deep learning methods of functional brain network classification are designed. The convolutional learning method learns the deep regional-connectivity features, while the recurrent learning method learns the deep adjacent positional features. The ELM (extreme learning machine)-boosted structure is also implemented to further improve the learning ability. Extensive experiments are conducted to evaluate and compare the AUC (area under curve), accuracy, recall, and training time of the proposed methods on a real-world dataset. Results indicate that (1) the proposed methods which learn deep features directly from brain networks outperform shallow learning methods and (2) models with the ELM-boosted structure achieve a higher performance. This paper explores the brain networks learning with deep features and ELM. The results demonstrate that the proposed methods provide a satisfactory learning ability in the application of AD detection.

CNNDLP: A Method Based on Convolutional Autoencoder and Convolutional Neural Network with Adjacent Edge Attention for Predicting lncRNA-Disease Associations.

It is well known that the unusual expression of long non-coding RNAs (lncRNAs) is closely related to the physiological and pathological processes of diseases. Therefore, inferring the potential lncRNA–disease associations are helpful for understanding the molecular pathogenesis of diseases. Most previous methods have concentrated on the construction of shallow learning models in order to predict lncRNA-disease associations, while they have failed to deeply integrate heterogeneous multi-source data and to learn the low-dimensional feature representations from these data. We propose a method based on the convolutional neural network with the attention mechanism and convolutional autoencoder for predicting candidate disease-related lncRNAs, and refer to it as CNNDLP. CNNDLP integrates multiple kinds of data from heterogeneous sources, including the associations, interactions, and similarities related to the lncRNAs, diseases, and miRNAs. Two different embedding layers are established by combining the diverse biological premises about the cases that the lncRNAs are likely to associate with the diseases. We construct a novel prediction model based on the convolutional neural network with attention mechanism and convolutional autoencoder to learn the attention and the low-dimensional network representations of the lncRNA–disease pairs from the embedding layers. The different adjacent edges among the lncRNA, miRNA, and disease nodes have different contributions for association prediction. Hence, an attention mechanism at the adjacent edge level is established, and the left side of the model learns the attention representation of a pair of lncRNA and disease. A new type of lncRNA similarity and a new type of disease similarity are calculated by incorporating the topological structures of multiple bipartite networks. The low-dimensional network representation of the lncRNA-disease pairs is further learned by the autoencoder based convolutional neutral network on the right side of the model. The cross-validation experimental results confirm that CNNDLP has superior prediction performance compared to the state-of-the-art methods. Case studies on stomach cancer, breast cancer, and prostate cancer further show the ability of CNNDLP for discovering the potential disease lncRNAs.

Data-Driven In-Situ Sonic-Log Synthesis in Shale Reservoirs for Geomechanical Characterization

SummaryCompressional–travel–time (DTC) and shear–travel–time (DTS) logs acquired using sonic–logging tools are crucial for subsurface geomechanical characterization. In this study, 13 “easy–to–acquire” conventional logs were processed using six shallow–regression–type supervised–learning models—namely, ordinary least squares (OLS), partial least squares (PLS), ElasticNet (EN), least absolute shrinkage and selection operator (LASSO), multivariate adaptive regression splines (MARS), and artificial–neural network (ANN)—to successfully synthesize DTC and DTS logs. Among the six models, ANN outperforms other models with R2 of 0.87 and 0.85 for the syntheses of DTC and DTS logs, respectively. The six shallow–learning models are trained and tested with 8,481 data points acquired from a 4,240–ft–depth interval of a shale reservoir in Well 1, and the trained models are deployed in Well 2 for purposes of blind testing against 2,920 data points from 1,460–ft–depth interval. After that, five clustering algorithms are applied on subsamples of 13 “easy–to–acquire” logs to identify clusters and compare them with the log–synthesis performance of the shallow–learning models. A dimensionality–reduction algorithm, t–distributed stochastic neighbor embedding (t–SNE), is used to visualize the petrophysical/statistical characteristics of the clustering algorithm. Hierarchical–clustering, density–based spatial clustering of application with noise (DBSCAN), and self–organizing–map (SOM) algorithms are sensitive to outliers and did not effectively differentiate the input data into consistent clusters. A Gaussian–mixture model can differentiate the various formations, but the clusters do not have a strong correlation with the performance of the log–synthesis models. Clusters identified using the K–means method have a strong correlation with the performance of the shallow–learning models. By combining the shallow–learning models for log synthesis with the K–means clustering, we propose a reliable workflow that can synthesize the DTC and DTS logs, as well as generate a reliability indicator for the synthesis logs to help the user better understand the performance of the shallow–learning models during deployment in new wells.

Empirical study of shallow and deep learning models for sarcasm detection using context in benchmark datasets

Sarcastic expressions tend to flip the polarity of posts when being analyzed for sentiments. Detecting sarcastic tone, which conveys a sharp, bitter, or cutting expression, remark or taunt in natural language is tricky even for humans, making its automated detection more arduous. Computational models for sarcasm detection have often relied on the content of utterances in isolation whereas using contextual information definitely improves it. This work is a preliminary to understand the what, how and why of using context in sentiment analysis. The concept of ‘context in use’ is described by exemplifying content-based local and global context to predict sarcasm in user-generated social textual data. In this research Twitter data of benchmark SemEval 2015 Task 11 and nearly 20 k posts from Reddit are classified as sarcastic or non-sarcastic using three predictive learning models. The first model is based on the conventional Term Frequency-Inverse Document Frequency (TF-IDF) weighting which is trained over three classifiers, namely the Multinomial Naive Bayes, Gradient Boosting and Random Forest and Ensemble Voting is utilized to generate the output. In the second model, a combination of semantic (sentiment) and pragmatic (punctuation) features are considered to model the context along with the top-200 TF-IDF features and results are observed using five baseline classifiers (Decision Tree, Support Vector Machine, Random Forest, K-Nearest Neighbour and Multi Layer Perceptron). The final model uses deep learning, that is, the Long–Short-Term-Memory (LSTM) and its variant Bi-directional LSTM applying GloVe (Global Vectors for Word Representation) for building semantic word embeddings and learning context. The empirical study using training set and test set performance metrics (Accuracy, Recall, Precision, F1 Score) is done to compare the three learning models for sarcasm classification within two datasets, and it is observed that Bi-directional LSTM model has the highest accuracy of 86.32% and 82.91% for the Twitter and Reddit datasets respectively.

Deep Learning for Monitoring of Human Gait: A Review

The essential human gait parameters are briefly reviewed, followed by a detailed review of the state of the art in deep learning for the human gait analysis. The modalities for capturing the gait data are grouped according to the sensing technology: video sequences, wearable sensors, and floor sensors, as well as the publicly available datasets. The established artificial neural network architectures for deep learning are reviewed for each group, and their performance are compared with particular emphasis on the spatiotemporal character of gait data and the motivation for multi-sensor, multi-modality fusion. It is shown that by most of the essential metrics, deep learning convolutional neural networks typically outperform shallow learning models. In the light of the discussed character of gait data, this is attributed to the possibility to extract the gait features automatically in deep learning as opposed to the shallow learning from the handcrafted gait features.

A deep separable neural network for human tissue identification in three-dimensional optical coherence tomography images

This research proposes a dilated depthwise separable network for human tissue identification using three-dimensional (3D) optical coherence tomography (OCT) images. Automatic human tissue identification has made it possible for fast pathological tissue analyses, detecting tissue changes over time, and efficiently making a precise therapy treatment plan. 3D medical image classification is a challenging task because of the indistinct tissue characteristics and computational efficiency. To address the challenge, a deep dilated depthwise separable convolutional neural network is proposed in this research. A depthwise separable architecture is introduced to improve parameter utilization efficiency. Dilated convolutions are applied to systematically aggregate multiscale contextual information and provide a large receptive field with a small number of trainable weights, which provide computational benefit. 2D convolutions are applied in the proposed model to enhance the computational efficiency. The constructed model is tested by performing a multi-class human thyroid tissue classification on 3D OCT images. For comparison, experimental results are obtained for texture feature-based shallow learning models and typical deep learning classification models. The results show that the proposed DDSCN model outperforms those state-of-art models and can improve accuracy by 3.2% compared with the best texture-based model and 2.27% compared with the best CNN model. The proposed deep model provides significant potential for the applicability of the deep learning technique to analyze medical images of human tissue while advancing the next generation of OCT-based real-time surgery image guidance.

Prediction of the crack condition of highway pavements using machine learning models

Departments of Transportation regularly evaluate the condition of pavements through visual inspections, nondestructive evaluations, image recognition models and learning algorithms. The above methodologies, though efficient, have drawn attention due to their subjective errors, uncertainties, noise effects and overfitting. To improve on the outcomes of the shallow learning models already used in pavement crack prediction, this paper reports on an investigation of the use of recursive partitioning and artificial neural networks (ANN; deep learning frameworks) in predicting the crack rating of pavements. Explanatory variables such as the average daily traffic and truck factor, roadway functional class, asphalt thickness, and pavement condition time series data are employed in the model formulation. Overall, it is observed that the recursive partitioning (regression tree – R2 > 0.8 and classification tree – R2 > 0.6) and ANN (continuous response – R2 > 0.8 and categorical response – R2 > 0.6) are compelling machine learning models for the prediction of the crack ratings based on their goodness-of-fit statistics, mean absolute deviation (MAD < 0.4) and the root mean square errors (RMSE between 0.30 and 0.65).

Accuracy Performance Degradation in Image Classification Models due to Concept Drift

Big data is playing a significant role in the current computing revolution. Industries and organizations are utilizing their insights for Business Intelligence by using Deep Learning Networks (DLN). However, dynamic characteristics of BD introduce many critical issues for DLN; Concept Drift (CD) is one of them. CD issue appears frequently in Online Supervised Learning environments in which data trends change over time. The problem may even worsen in a BD environment due to the veracity and variability factors. The CD issue may render the DLN inapplicable by degrading the accuracy of classification results in DLN which is a very serious issue that needs to be addressed. Therefore, these DLN need to quickly adapt to changes for maintaining the accuracy level of the results. To overcome classification accuracy, we need some dynamical changes in the existing DLN. Therefore, in this paper, we examine some of the existing Shallow Learning and Deep Learning models and their behavior before and after the Concept Drift (in experiment 1) and validate the pre-trained Deep Learning network (ResNet-50). In future work, this experiment will examine the most recent pre-trained DLN (Alex Net, VGG16, VGG19) and identify their suitability to overcome Concept Drift using fine-tuning and transfer learning approaches.

Marginal Deep Architecture: Stacking Feature Learning Modules to Build Deep Learning Models

Recently, many deep models have been proposed in different fields, such as image classification, object detection, and speech recognition. However, most of these architectures require a large amount of training data and employ random initialization. In this paper, we propose to stack feature learning modules for the design of deep architectures. Specifically, marginal Fisher analysis (MFA) is stacked layer-by-layer for the initialization and we call the constructed deep architecture marginal deep architecture (MDA). When implementing the MDA, the weight matrices of MFA are updated layer-by-layer, which is a supervised pre-training method and does not need a large scale of data. In addition, several deep learning techniques are applied to this architecture, such as backpropagation, dropout, and denoising, to fine-tune the model. We have compared MDA with some feature learning and deep learning models on several practical applications, such as handwritten digits recognition, speech recognition, historical document understanding, and action recognition. The extensive experiments show that the performance of MDA is better than not only shallow feature learning models but also related deep learning models in these tasks.

Accurate EEG-Based Emotion Recognition on Combined Features Using Deep Convolutional Neural Networks

In order to improve the accuracy of emotional recognition by end-to-end automatic learning of emotional features in spatial and temporal dimensions of electroencephalogram (EEG), an EEG emotional feature learning and classification method using deep convolution neural network (CNN) was proposed based on temporal features, frequential features, and their combinations of EEG signals in DEAP dataset. The shallow machine learning models including bagging tree (BT), support vector machine (SVM), linear discriminant analysis (LDA), and Bayesian linear discriminant analysis (BLDA) models and deep CNN models were used to make emotional binary classification experiments on DEAP datasets in valence and arousal dimensions. The experimental results showed that the deep CNN models which require no feature engineering achieved the best recognition performance on temporal and frequency combined features in both valence and arousal dimensions, which is 3.58% higher than the performance of the best traditional BT classifier in valence dimension and 3.29% higher than that of BT classifier in arousal dimension.

AUTOMATIC DETECTION OF CARDIOVASCULAR DISEASE USING DEEP KERNEL EXTREME LEARNING MACHINE

The electrocardiogram (ECG) is a principal signal employed to automatically diagnose cardiovascular disease in shallow and deep learning models. However, ECG feature extraction is required and this may reduce diagnosis accuracy in traditional shallow learning models, while backward propagation (BP) algorithm used by the traditional deep learning models has the disadvantages of local minimization and slow convergence rate. To solve these problems, a new deep learning algorithm called deep kernel extreme learning machine (DKELM) is proposed by combining the extreme learning machine auto-encoder (ELM-AE) and kernel ELM (KELM). In the new DKELM architecture with [Formula: see text] hidden layers, ELM-AEs are employed by the front [Formula: see text] hidden layers for feature extraction in the unsupervised learning process, which can effectively extract abstract features from the original ECG signal. To overcome the “dimension disaster” problem, the kernel function is introduced into ELM to act as classifier by the [Formula: see text]th hidden layer in the supervised learning process. The experiments demonstrate that DKELM outperforms the BP neural network, support vector machine (SVM), extreme learning machine (ELM), deep auto-encoder (DAE), deep belief network (DBN) in classification accuracy. Though the accuracy of convolutional neural network (CNN) is almost the same as DKELM, the computing time of CNN is much longer than DKELM.

Deep generative learning for automated EHR diagnosis of traditional Chinese medicine

BackgroundComputer-aided medical decision-making (CAMDM) is the method to utilize massive EMR data as both empirical and evidence support for the decision procedure of healthcare activities. Well-developed information infrastructure, such as hospital information systems and disease surveillance systems, provides abundant data for CAMDM. However, the complexity of EMR data with abstract medical knowledge makes the conventional model incompetent for the analysis. Thus a deep belief networks (DBN) based model is proposed to simulate the information analysis and decision-making procedure in medical practice. The purpose of this paper is to evaluate a deep learning architecture as an effective solution for CAMDM. MethodsA two-step model is applied in our study. At the first step, an optimized seven-layer deep belief network (DBN) is applied as an unsupervised learning algorithm to perform model training to acquire feature representation. Then a support vector machine model is adopted to DBN at the second step of the supervised learning. There are two data sets used in the experiments. One is a plain text data set indexed by medical experts. The other is a structured dataset on primary hypertension. The data are randomly divided to generate the training set for the unsupervised learning and the testing set for the supervised learning. The model performance is evaluated by the statistics of mean and variance, the average precision and coverage on the data sets. Two conventional shallow models (support vector machine / SVM and decision tree / DT) are applied as the comparisons to show the superiority of our proposed approach. ResultsThe deep learning (DBN + SVM) model outperforms simple SVM and DT on two data sets in terms of all the evaluation measures, which confirms our motivation that the deep model is good at capturing the key features with less dependence when the index is built up by manpower. ConclusionsOur study shows the two-step deep learning model achieves high performance for medical information retrieval over the conventional shallow models. It is able to capture the features of both plain text and the highly-structured database of EMR data. The performance of the deep model is superior to the conventional shallow learning models such as SVM and DT. It is an appropriate knowledge-learning model for information retrieval of EMR system. Therefore, deep learning provides a good solution to improve the performance of CAMDM systems.

When Low Rank Representation Based Hyperspectral Imagery Classification Meets Segmented Stacked Denoising Auto-Encoder Based Spatial-Spectral Feature

When confronted with limited labelled samples, most studies adopt an unsupervised feature learning scheme and incorporate the extracted features into a traditional classifier (e.g., support vector machine, SVM) to deal with hyperspectral imagery classification. However, these methods have limitations in generalizing well in challenging cases due to the limited representative capacity of the shallow feature learning model, as well as the insufficient robustness of the classifier which only depends on the supervision of labelled samples. To address these two problems simultaneously, we present an effective low-rank representation-based classification framework for hyperspectral imagery. In particular, a novel unsupervised segmented stacked denoising auto-encoder-based feature learning model is proposed to depict the spatial-spectral characteristics of each pixel in the imagery with deep hierarchical structure. With the extracted features, a low-rank representation based robust classifier is then developed which takes advantage of both the supervision provided by labelled samples and unsupervised correlation (e.g., intra-class similarity and inter-class dissimilarity, etc.) among those unlabelled samples. Both the deep unsupervised feature learning and the robust classifier benefit, improving the classification accuracy with limited labelled samples. Extensive experiments on hyperspectral imagery classification demonstrate the effectiveness of the proposed framework.

An End-to-End Model Based on Improved Adaptive Deep Belief Network and Its Application to Bearing Fault Diagnosis

Effective machinery prognostics and health management play a crucial role in ensuring the safe and continuous operation of equipment, and satisfactory characteristics' expression of machine health status plays a key role in the ability to diagnose faults with high accuracy. At present, most methods based on signal processing and the shallow learning model rely on artificial feature extraction to identify the machine fault type. In practical applications, however, meaningful health management requires correct recognition of not only the health type but also the fault degree, if any occurs. Such recognition is useful for determining the priority level of mechanical maintenance and minimizing economic losses. Deep learning techniques, such as deep belief network (DBN), have demonstrated great potential in exploring characteristic information from machine status signals. In this paper, an end-to-end fault diagnosis model based on an adaptive DBN optimized by the Nesterov moment (NM) is proposed to extract deep representative features from rotating machinery and recognize bearing fault types and degrees simultaneously. Frequency-domain signals are inputted into the model for feature learning, and NM is introduced to the training process of the DBN model. Individual adaptive learning rate algorithms are then applied to optimize parameter updating. The performance of the proposed method is validated using a self-made bearing fault test platform, and the model is shown to achieve satisfactory convergence and a testing accuracy higher than those obtained from standard DBN and support vector machine.

Comparative Study of Shallow Learning Models for Generating Compressional and Shear Traveltime Logs

Comparative Study of Shallow Learning Models for Generating Compressional and Shear Traveltime Logs

Deep Neural Networks for Learning Spatio-Temporal Features From Tomography Sensors

We demonstrate accurate spatio-temporal gait data classification from raw tomography sensor data without the need to reconstruct images. This is based on a simple yet efficient machine learning methodology based on a convolutional neural network architecture for learning spatio-temporal features, automatically end-to-end from raw sensor data. In a case study on a floor pressure tomography sensor, experimental results show an effective gait pattern classification F -score performance of 97.88 $\pm$ 1.70%. It is shown that the automatic extraction of classification features from raw data leads to a substantially better performance, compared to features derived by shallow machine learning models that use the reconstructed images as input, implying that for the purpose of automatic decision-making it is possible to eliminate the image reconstruction step. This approach is portable across a range of industrial tasks that involve tomography sensors. The proposed learning architecture is computationally efficient, has a low number of parameters and is able to achieve reliable classification F -score performance from a limited set of experimental samples. We also introduce a floor sensor dataset of 892 samples, encompassing experiments of 10 manners of walking and 3 cognitive-oriented tasks to yield a total of 13 types of gait patterns.

Shallow learning model for diagnosing neuro muscular disorder from splicing variants

PurposeDiagnosing genetic neuromuscular disorder such as muscular dystrophy is complicated when the imperfection occurs while splicing. This paper aims in predicting the type of muscular dystrophy from the gene sequences by extracting the well-defined descriptors related to splicing mutations. An automatic model is built to classify the disease through pattern recognition techniques coded in python using scikit-learn framework.Design/methodology/approachIn this paper, the cloned gene sequences are synthesized based on the mutation position and its location on the chromosome by using the positional cloning approach. For instance, in the human gene mutational database (HGMD), the mutational information for splicing mutation is specified as IVS1-5 T > G indicates (IVS - intervening sequence or introns), first intron and five nucleotides before the consensus intron site AG, where the variant occurs in nucleotide G altered to T. IVS (+ve) denotes forward strand 3′– positive numbers from G of donor site invariant and IVS (−ve) denotes backward strand 5′ – negative numbers starting from G of acceptor site. The key idea in this paper is to spot out discriminative descriptors from diseased gene sequences based on splicing variants and to provide an effective machine learning solution for predicting the type of muscular dystrophy disease with the splicing mutations. Multi-class classification is worked out through data modeling of gene sequences. The synthetic mutational gene sequences are created, as the diseased gene sequences are not readily obtainable for this intricate disease. Positional cloning approach supports in generating disease gene sequences based on mutational information acquired from HGMD. SNP-, gene- and exon-based discriminative features are identified and used to train the model. An eminent muscular dystrophy disease prediction model is built using supervised learning techniques in scikit-learn environment. The data frame is built with the extracted features as numpy array. The data are normalized by transforming the feature values into the range between 0 and 1 aid in scaling the input attributes for a model. Naïve Bayes, decision tree, K-nearest neighbor and SVM learned models are developed using python library framework in scikit-learn.FindingsTo the best knowledge of authors, this is the foremost pattern recognition model, to classify muscular dystrophy disease pertaining to splicing mutations. Certain essential SNP-, gene- and exon-based descriptors related to splicing mutations are proposed and extracted from the cloned gene sequences. An eminent model is built using statistical learning technique through scikit-learn in the anaconda framework. This paper also deliberates the results of statistical learning carried out with the same set of gene sequences with synonymous and non-synonymous mutational descriptors.Research limitations/implicationsThe data frame is built with the Numpy array. Normalizing the data by transforming the feature values into the range between 0 and 1 aid in scaling the input attributes for a model. Naïve Bayes, decision tree, K-nearest neighbor and SVM learned models are developed using python library framework in scikit-learn. While learning the SVM model, the cost, gamma and kernel parameters are tuned to attain good results. Scoring parameters of the classifiers are evaluated using tenfold cross-validation using metric functions of scikit-learn library. Results of the disease identification model based on non-synonymous, synonymous and splicing mutations were analyzed.Practical implicationsCertain essential SNP-, gene- and exon-based descriptors related to splicing mutations are proposed and extracted from the cloned gene sequences. An eminent model is built using statistical learning technique through scikit-learn in the anaconda framework. The performance of the classifiers are increased by using different estimators from the scikit-learn library. Several types of mutations such as missense, non-sense and silent mutations are also considered to build models through statistical learning technique and their results are analyzed.Originality/valueTo the best knowledge of authors, this is the foremost pattern recognition model, to classify muscular dystrophy disease pertaining to splicing mutations.

Handwritten isolated Bangla compound character recognition: A new benchmark using a novel deep learning approach

In this work, a novel deep learning technique for the recognition of handwritten Bangla isolated compound character is presented and a new benchmarkof recognition accuracy on the CMATERdb 3.1.3.3 dataset is reported. Greedy layer wise training of Deep Neural Network has helped to make significant strides in various pattern recognition problems. We employ layerwise training to Deep Convolutional Neural Networks (DCNN) in a supervised fashion and augment the training process with the RMSProp algorithm to achieve faster convergence. We compare results with those obtained from standard shallow learning methods with predefined features, as well as standard DCNNs. Supervised layerwise trained DCNNs are found to outperform standard shallow learning models such as Support Vector Machines as well as regular DCNNs of similar architecture by achieving error rate of 9.67% thereby setting a new benchmark on the CMATERdb 3.1.3.3 with recognition accuracy of 90.33%, representing an improvement of nearly 10%.

Classification of Cytochrome P450 1A2 Inhibitors and Noninhibitors Based on Deep Belief Network

The cytochrome P450 (CYP) superfamily, exists in the human liver, is responsible for more than 90% of the metabolism of clinical drugs. So it is necessary to adopt a new kind of computer simulation methods that can predict the rejection capability of compounds for a concrete CYPs isoform. In this work, a model is presented for classification of CYP450 1A2 inhibitors and noninhibitors based on a multi-tiered deep belief network (DBN) on a large dataset. The dataset composed of more than 13,000 heterogeneous compounds was acquired from PubChem. Firstly, 139 2D and 53 3D descriptors are calculated and preprocessed. Then, the unsupervised learning method is used to train DBN model to automatically extract multiple levels of distributed representation from the descriptors of training set. Finally, by using testing set and external validation set, we evaluate the classified performance of DBN for the inhibition of CYP1A2. Meanwhile, the proposed model is compared with shallow machine learning models (support vector machine (SVM) and artificial neural network (ANN)). We also discussed the performance of DBN by comparing it with different features combination. The experimental results showed that DBN has a better prediction ability compared with SVM and ANN. And these models combined with the features of 2D and 3D obtain the best forecast accuracy.

Heterogeneous visual features integration for image recognition optimization in internet of things

Recently, a large number of physical devices, together with distributed information systems, deployed in internet of things (IoT), are collecting more and more images. Such collected images recognition poses an important challenge on optimization in internet of things. Specially, most of existing methods only adopt shallow learning models to integrate various features of images for recognition limiting classification accuracy. In this paper, we propose a multimodal deep learning (MMDL) approach to integrate heterogeneous visual features by considering each type of visual feature as one modality for image recognition optimization in internet of things. In our scheme, we extract the high-level abstraction of each modality by a stacked autoencoders. Furthermore, we design a back propagation algorithm with shared weights learned from a softmax layer to update the pretrained parameters of multiple stacked autoencoders simultaneously. The integration is performed by concatenating the last hidden layers of the multimodal stacked autoencoders architecture. Extensive experiments are carried out on three datasets i.e. Animal with Attributes, NUS-WIDE-OBJECT, and Handwritten Numerals, by comparison with SVM, SAE, and AMMSS. Results demonstrate that our scheme has superior performance on heterogeneous visual features integration for image recognition optimization in internet of things.