Latent Space Representation Research Articles

Classification of electromyographic hand gesture signals using machine learning techniques

The electromyogram (EMG) signals from an individual’s muscles can reflect the biomechanics of human movement. The accurate classification of individual and combined finger movements using surface EMG signals is able to support many applications such as dexterous prosthetic hand control. The existing research of EMG-based hand gesture classification faces the challenges of inaccurate classification, insufficient generalization ability and weak robustness. To address these problems, this paper proposes a deep learning model that combines convolutional auto-encoder and convolutional neural network (CAE+CNN) to classify an EMG dataset consisting of 10 classes of hand gestures. The proposed method shrinks the inputs into a smaller latent space representation using CAE and the resultant compressed features are served as inputs of CNN, which reduces the redundancy of EMG signals and improves the classification accuracy and training efficiency. Besides, to enhance the robustness and generalization ability for classification, a data processing approach is proposed which combines the windowing method and majority voting of the obtained results from the classifier. In addition, comprehensive comparative study is carried out with 8 widely applied and state-of-the-art classifiers in terms of classification accuracy, robustness subject to noise and statistical analysis (sensitivity, specificity, precision, F1 Score and Matthews correlation coefficient). The results demonstrates that the integration of windowing method, CAE+CNN and majority voting achieves the best performance (99.38% test accuracy for the data without adding noise, which is 3.78% higher than the best classifier used for comparison), strongest robustness (achieved 98.13% test accuracy when Gaussian noise of level 1e-5 is added to the raw dataset, which is 4.07% higher than the best classifier used for comparison) and statistical properties compared to other classifiers, which shows the potential for healthcare applications such as movement intention detection and dexterous prostheses control.

Adversarial autoencoder for visualization and classification of human activity: Application to a low-cost commercial force plate

The ability to visualize and interpret high dimensional time-series data will be critical as wearable and other sensors are adopted in rehabilitation protocols. This study proposes a latent space representation of high dimensional time-series data for data visualization. For that purpose, a deep learning model called Adversarial AutoEncoder (AAE) is proposed to perform efficient data dimensionality reduction by considering unsupervised and semi-supervised adversarial training.Eighteen subjects were recruited for the experiment and performed two sets of exercises (upper and lower body) on the Wii Balance Board. Then, the accuracy of the latent space representation is evaluated on both sets of exercises separately. Data dimensionality reduction with conventional Machine Learning (ML) and supervised Deep Learning (DL) classification are also performed to compare the efficiency of AAE approaches.The results showed that AAE can outperform conventional ML approaches while providing close results to DL supervised classification.AAE approaches for data visualization are a promising approach to monitor the subject’s movements and detect adverse events or similarity with previous data, providing an intuitive way to monitor the patient’s progress and provide potential information for rehabilitation tracking.

Time-dependent reliability prediction using transfer learning

One of the key challenges in reliability estimation is the acquisition of failure information especially under real-life scenarios or computationally expensive long-run simulations. Due to the lack of failure information, it becomes a challenging task to estimate time-dependent reliability of a structure or component. In this paper, a preliminary study of time-dependent reliability prediction using transfer learning is given by utilizing only current time performance function information. Transfer learning, specifically domain adaptation, is used in this work in order to find a representation in latent space where the statistical properties of nonstationary stochastic processes such as variance are preserved and their distribution parameters at different times are made close to each other. Correlated random samples of the stochastic processes are generated in present and future time intervals in order to find common latent space using domain adaptation via transfer component analysis (TCA). Due to the excellent computational efficiency, a Kriging surrogate model is built in the present interval only using an adaptive sampling strategy. The reliability in the future is predicted using the same surrogate model without retraining it using future performance function information. Monte Carlo simulation (MCS) is used to validate the results of the proposed method. The results show that the proposed method can predict the failure probability of time-dependent problems efficiently with significant accuracy.

Time-series learning of latent-space dynamics for reduced-order model closure

We study the performance of long short-term memory networks (LSTMs) and neural ordinary differential equations (NODEs) in learning latent-space representations of dynamical equations for an advection-dominated problem given by the viscous Burgers equation. Our formulation is devised in a nonintrusive manner with an equation-free evolution of dynamics in a reduced space with the latter being obtained through a proper orthogonal decomposition. In addition, we leverage the sequential nature of learning for both LSTMs and NODEs to demonstrate their capability for closure in systems that are not completely resolved in the reduced space. We assess our hypothesis for two advection-dominated problems given by the viscous Burgers equation. We observe that both LSTMs and NODEs are able to reproduce the effects of the absent scales for our test cases more effectively than does intrusive dynamics evolution through a Galerkin projection. This result empirically suggests that time-series learning techniques implicitly leverage a memory kernel for coarse-grained system closure as is suggested through the Mori–Zwanzig formalism.

Adaptive Adversarial Latent Space for Novelty Detection

Novelty detection is a challenging task of identifying whether a new sample obeys to a known class. Note that the boundary between normal and novel is not clear enough in existing works, resulting from adequately reconstructing novel samples or crudely reconstructing normal samples. To tackle the above issues, we propose a general framework named Adaptive Adversarial Latent Space (AALS), which mainly consists of two components, Adaptive Latent Space Generator (ALSG) and Constrained-based Adversarial AutoEncoder (CAAE). ALSG is established to obtain the real latent space distribution by adaptive mapper and the manner of adversarial learning. Moreover, CAAE is presented to obtain a better boundary between normal and novel by Global and Local Channel-wise Attention (GLCA), which is proposed to reweight the global and local information in different channels. Given a sample, the latent space representation <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$z$ </tex-math></inline-formula> of the sample is firstly obtained by encoder with GLCA. Next, the output of ALSG is applied to constrain the latent space by latent adversarial loss. Further, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$z$ </tex-math></inline-formula> is transformed into the decoder to reconstruct the given sample with GLCA. Experiments on three available datasets demonstrates the effectiveness of the proposed method, which achieves state-of-the-art performance compared with other methods.

Video Popularity Prediction: An Autoencoder Approach With Clustering

Autoencoders implemented by artificial neural networks (ANNs) are utilized to learn the latent space representation of data in an unsupervised manner, and they have been widely used in recommender systems. For instance, several collaborative denoising autoencoder (CDAE) models have shown that their performance gains outperform that of the collaborative filtering based (CF-based) models. In this work, a near-optimal Top- $K$ forecasting solution is proposed for our advanced autoencoder recommender systems. We propose a method which utilizes CDAE model in predicting the Top- $K$ popular videos in an upcoming time period. In order to improve the prediction accuracy, we also propose an autoencoder based recommendation algorithm with the help of $K$ -means clustering that upgrades the performance of the original autoencoder model. The experimental results show that our method increases significantly the Average Precision (AP) and Recall values by nearly 30%. We then further utilize our proposed autoencoder model with clustering in predicting Top- $K$ popular videos. The applications of predicting Top- $K$ popular videos can be used in the video delivery for the Mobile Edge Computing (MEC) environment to avoid bottleneck in the constricted capacity of backhaul link. Namely, the performance gain will be upgraded if our proposed method precisely predicts and caches the Top- $K$ popular videos in advance with the help of a better forecasting model.

LPD-AE: Latent Space Representation of Large-Scale 3D Point Cloud

The effective latent space representation of point cloud provides a foremost and fundamental manner that can be used for challenging tasks, including point cloud based place recognition and reconstruction, especially in large-scale dynamic environments. In this paper, we present a novel deep neural network, LPD-AE(Large-scale Place Description AutoEncoder Network), to obtain meaningful local and contextual features for the generation of latent space from 3D point cloud directly. The encoder network constructs the discriminative global descriptors to realize high accuracy and robust place recognition, which contributed by extracting the local neighbor geometric features and aggregating neighborhood relationships both in feature space and physical space. The decoder network performs hierarchical reconstruction on coarse key points and ultimately produce dense point clouds, which shows that it is capable of reconstructing a full point cloud frame from a single compact but high dimensional descriptor. Our proposed network demonstrates performance that is comparable to the state-of-the-art approaches. With the benefit of the LPD-AE, many computationally complex tasks that rely directly on point clouds can be effortlessly conducted on latent space with lower memory costs, such as relocalization, loop closure detection, and map compression reconstruction. Comprehensive validations on Oxford RobotCar dataset, KITTI dataset, and our freshly collected dataset, which contains multiple trials of repeated routes in different weather and at different times, manifest its potency for real robotic and self-driving implementation. The source code is available at https://github.com/Suoivy/LPD-AE .

Deciphering protein evolution and fitness landscapes with latent space models

Protein sequences contain rich information about protein evolution, fitness landscapes, and stability. Here we investigate how latent space models trained using variational auto-encoders can infer these properties from sequences. Using both simulated and real sequences, we show that the low dimensional latent space representation of sequences, calculated using the encoder model, captures both evolutionary and ancestral relationships between sequences. Together with experimental fitness data and Gaussian process regression, the latent space representation also enables learning the protein fitness landscape in a continuous low dimensional space. Moreover, the model is also useful in predicting protein mutational stability landscapes and quantifying the importance of stability in shaping protein evolution. Overall, we illustrate that the latent space models learned using variational auto-encoders provide a mechanism for exploration of the rich data contained in protein sequences regarding evolution, fitness and stability and hence are well-suited to help guide protein engineering efforts.

Unsupervised feature selection via adaptive hypergraph regularized latent representation learning

Due to the rapid development of multimedia technology, a large number of unlabelled data with high dimensionality need to be processed. The high dimensionality of data not only increases the computation burden of computer hardware, but also hinders algorithms to obtain optimal performance. Unsupervised feature selection, which is regarded as a means of dimensionality reduction, has been widely recognized as an important and challenging pre-step for many machine learning and data mining tasks. However, we observe that there are at least two issues in previous unsupervised feature selection methods. Firstly, traditional unsupervised feature selection algorithms usually assume that the data instances are identically distributed and there is no dependency between them. However, the data instances are not only associated with high dimensional features but also inherently interconnected with each other. Secondly, the traditional similarity graph used in previous methods can only describe the pair-wise relations of data, but cannot capture the high-order relations, so that the complex structures implied in the data cannot be sufficiently exploited. In this work, we propose a robust unsupervised feature selection method which embeds the latent representation learning into feature selection. Instead of measuring the feature importances in original data space, the feature selection is carried out in the learned latent representation space which is more robust to noises. In order to capture the local manifold geometrical structure of original data in a high-order manner, a hypergraph is adaptively learned and embedded into the resultant model. An efficient alternating algorithm is developed to optimize the problem. Experimental results on eight benchmark data sets demonstrate the effectiveness of the proposed method.

Speed Adaptation in Learning from Demonstration through Latent Space Formulation

SUMMARYPerforming actions in a timely manner is an indispensable aspect in everyday human activities. Accordingly, it has to be present in robotic systems if they are going to seamlessly interact with humans. The current work addresses the problem of learning both the spatial and temporal characteristics of human motions from observation. We formulate learning as a mapping between two worlds (the observed and the action ones). This mapping is realized via an abstract intermediate representation termed “Latent Space.” Learned actions can be subsequently invoked in the context of more complex human–robot interaction (HRI) scenarios. Unlike previous learning from demonstration (LfD) methods that cope only with the spatial features of an action, the formulated scheme effectively encompasses spatial and temporal aspects. Learned actions are reproduced under the high-level control of a time-informed task planner. During the implementation of the studied scenarios, temporal and physical constraints may impose speed adaptations in the reproduced actions. The employed latent space representation readily supports such variations, giving rise to novel actions in the temporal domain. Experimental results demonstrate the effectiveness of the proposed scheme in the implementation of HRI scenarios. Finally, a set of well-defined evaluation metrics are introduced to assess the validity of the proposed approach considering the temporal and spatial consistency of the reproduced behaviors.

Inverse Design of Solid-State Materials via a Continuous Representation

Summary The non-serendipitous discovery of materials with targeted properties is the ultimate goal of materials research, but to date, materials design lacks the incorporation of all available knowledge to plan the synthesis of the next material. This work presents a framework for learning a continuous representation of materials and building a model for new discovery using latent space representation. The ability of autoencoders to generate experimental materials is demonstrated with vanadium oxides via rediscovery of experimentally known structures when the model was trained without them. Approximately 20,000 hypothetical materials are generated, leading to several completely new metastable VxOy materials that may be synthesizable. Comparison with genetic algorithms suggests computational efficiency of generative models that can explore chemical compositional space effectively by learning the distributions of known materials for crystal structure prediction. These results are an important step toward machine-learned inverse design of inorganic functional materials using generative models.

Video anomaly detection and localization via multivariate gaussian fully convolution adversarial autoencoder

In this paper, we present a novel deep learning based method for video anomaly detection and localization. The key idea of our approach is that the latent space representations of normal samples are trained to accord with a specific prior distribution by the proposed deep neural network - Multivariate Gaussian Fully Convolution Adversarial Autoencoder (MGFC-AAE), while the latent representations of anomalies do not. In order to extract deep features from input samples as latent representations, a convolutional neural network (CNN) is employed for the encoder of the deep network. Based on the probability that the test sample is associated with the prior distribution, an energy-based method is applied to obtain its anomaly score. A two-stream framework is utilized to integrate the appearance and motion cues to achieve more comprehensive detection results, taking the gradient and optical flow patches as inputs for each stream. Besides, a multi-scale patch structure is put forward to handle the perspective of some video scenes. Experiments are conducted on three public datasets, results verify that our framework can accurately detect and locate abnormal objects in various video scenes, achieving competitive performance when compared with other state-of-the-art works.

Improved Gaussian–Bernoulli restricted Boltzmann machine for learning discriminative representations

Restricted Boltzmann machines (RBMs) have received considerable research interest in recent years because of their capability to discover latent representations in an unsupervised manner. The standard RBM is only suitable for processing binary-valued data. To address this limitation, the Gaussian–Bernoulli RBM (GRBM) has been designed to model real-valued data, particularly images. A GRBM that seeks to map real-valued data nonlinearly into a latent representation space is typically trained by contrastive divergence learning. However, most existing GRBM-based models neglect the inherent interpoint affinity information from the original data that can be used to enhance the expression ability of the model. In this study, a novel interpoint-affinity-based GRBM (abGRBM) is proposed to learn discriminative representations in the hidden layer. By incorporating the interpoint affinity information into the training process of the GRBM, the proposed model can not only utilize the GRBM’s powerful latent representation learning capabilities for real-valued data, it can also transform the original data into another space with improved separability. We prove the availability of our model using several image datasets of Microsoft Research Asia Multimedia for unsupervised clustering and supervised classification tasks. The experimental results show the superior performance of the GRBM in discovering discriminative representations and demonstrate the effectiveness of the affinity information.

Predicting the Spread of Malware Outbreaks Using Autoencoder Based Neutral Networks

Malware Outbreaks are pervasive in today's digital world. However, there is a lack of awareness on part of general public on how to safeguard against such attacks and a need for increased cooperation between various national and international research as well as governmental organizations to combat the threat. On the positive side, cyber security websites, blogs and newsletters post articles outlining the working and spread of a malware outbreak and steps to recover from the same as well. In this project, an effective approach to predicting the spread of malware outbreaks is presented. The scope of the project is 15 Malware Outbreaks and the approach involves collecting these cyber aware articles from the web, assigning them to the 15 Malware Outbreaks using Topic Modeling and Similarity Analysis and along with Spread information of the Malware Outbreaks, this is input to auto encoder neural network for learning latent space representations which are further used to predict the spread of malware outbreak as either high or low spread outbreak, achieving a prediction accuracy of 75.56. This work can be used to process large amount of cyber aware content for effective and accurate prediction in the era of much-needed cyber security.

Unsupervised feature selection via latent representation learning and manifold regularization

With the rapid development of multimedia technology, massive unlabelled data with high dimensionality need to be processed. As a means of dimensionality reduction, unsupervised feature selection has been widely recognized as an important and challenging pre-step for many machine learning and data mining tasks. Traditional unsupervised feature selection algorithms usually assume that the data instances are identically distributed and there is no dependency between them. However, the data instances are not only associated with high dimensional features but also inherently interconnected with each other. Furthermore, the inevitable noises mixed in data could degenerate the performances of previous methods which perform feature selection in original data space. Without label information, the connection information between data instances can be exploited and could help select relevant features. In this work, we propose a robust unsupervised feature selection method which embeds the latent representation learning into feature selection. Instead of measuring the feature importances in original data space, the feature selection is carried out in the learned latent representation space which is more robust to noises. The latent representation is modelled by non-negative matrix factorization of the affinity matrix which explicitly reflects the relationships of data instances. Meanwhile, the local manifold structure of original data space is preserved by a graph based manifold regularization term in the transformed feature space. An efficient alternating algorithm is developed to optimize the proposed model. Experimental results on eight benchmark datasets demonstrate the effectiveness of the proposed method.

Decomposing Cell Identity for Transfer Learning across Cellular Measurements, Platforms, Tissues, and Species.

Analysis of gene expression in single cells allows for decomposition of cellular states as low-dimensional latent spaces. However, the interpretation and validation of these spaces remains a challenge. Here, we present scCoGAPS, which defines latent spaces from a source single-cell RNA-sequencing (scRNA-seq) dataset, and projectR, which evaluates these latent spaces in independent target datasets via transfer learning. Application of developing mouse retina to scRNA-Seq reveals intrinsic relationships across biological contexts and assays while avoiding batch effects and other technical features. We compare the dimensions learned in this source dataset to adult mouse retina, a time-course of human retinal development, select scRNA-seq datasets from developing brain, chromatin accessibility data, and a murine-cell type atlas to identify shared biological features. These tools lay the groundwork for exploratory analysis of scRNA-seq data via latent space representations, enabling a shift in how we compare and identify cells beyond reliance on marker genes or ensemble molecular identity.

Do Stocks Stalk Other Stocks in Their Complex Network? A Complex Networks Approach to Stock Market Dynamics

This article presents a complex network approach to analyze the dynamics of the US stock market. The authors first introduce a machine learning technique to learn a latent space representation of the financial instruments using a complex networks approach to analyze the dynamics of the stock market. They show that the clustering of co-moving assets and breaking of already existing clusters have interesting connections to regime changes in the market. The authors also present some preliminary ideas on incorporating the effects of news and announcements in the dynamics of the assets.

Latent Representation Learning for Alzheimer's Disease Diagnosis With Incomplete Multi-Modality Neuroimaging and Genetic Data.

The fusion of complementary information contained in multi-modality data [e.g., magnetic resonance imaging (MRI), positron emission tomography (PET), and genetic data] has advanced the progress of automated Alzheimer's disease (AD) diagnosis. However, multi-modality based AD diagnostic models are often hindered by the missing data, i.e., not all the subjects have complete multi-modality data. One simple solution used by many previous studies is to discard samples with missing modalities. However, this significantly reduces the number of training samples, thus leading to a sub-optimal classification model. Furthermore, when building the classification model, most existing methods simply concatenate features from different modalities into a single feature vector without considering their underlying associations. As features from different modalities are often closely related (e.g., MRI and PET features are extracted from the same brain region), utilizing their inter-modality associations may improve the robustness of the diagnostic model. To this end, we propose a novel latent representation learning method for multi-modality based AD diagnosis. Specifically, we use all the available samples (including samples with incomplete modality data) to learn a latent representation space. Within this space, we not only use samples with complete multi-modality data to learn a common latent representation, but also use samples with incomplete multi-modality data to learn independent modality-specific latent representations. We then project the latent representations to the label space for AD diagnosis. We perform experiments using 737 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, and the experimental results verify the effectiveness of our proposed method.

Constrained matrix factorization for semi-weakly learning with label proportions

Learning with label proportions (LLP) is a weakly supervised learning problem that is conceivable in many real-world applications, where the training data is given in bags of instances, and only knowing the proportions of data points belonging to a particular category for each bag. However, how to effectively address the LLP problem with high dimensional data and some labeled samples is still a challenging problem. In this paper, we firstly propose a novel learning problem called semi-weakly learning with label proportions (SLLP), which has more extensive application scenarios. Then, we contribute a novel method based on non-negative matrix factorization, called Proportion Constrained Matrix Factorization (PCMF). It can not only effectively incorporate the label and proportion information, but also explore the local manifold structure information of training data. Moreover, the proposed method can make the data points from the same class be more likely merged together in the latent representation space, which leads to the more discriminating power. Sufficient experimental results on the benchmark datasets demonstrates its superiority over the state-of-the-art methods for LLP problem and efficiency on solving the SLLP problem.

Learning Robust Auto-Encoders With Regularizer for Linearity and Sparsity

Unsupervised feature learning via auto-encoders results in low-dimensional representations in latent space that capture the patterns of input data. The auto-encoders with robust regularization learn qualified features that are less sensitive to small perturbations of inputs. However, the previous robust auto-encoders highly depend on pre-defined structure settings and often learn full-connected networks that are easily prone to over-fitting. To solve the above limitations, we propose in this paper an explicitly regularized framework which improves the sparsity and flexibility of robust auto-encoders. First, our model encourages the activation functions to automatically adjust themselves between linear and non-linear ones. Second, the mapping functions of the encoder are constrained by group sparsity and exclusive sparsity to reduce the redundancy of parameters. The proximal gradient method is used to optimize our model since the objective function contains non-smooth components. We conduct experiments in single-layer and multiple-layer auto-encoders in the classification task. The numerical results show that our model achieves better accuracy than baseline models. Our method also shows better performance in denoising task.