Unsupervised Deep Learning Technique Research Articles

High-resolution extracellular pH imaging of liver cancer with multiparametric MR using Deep Image Prior.

Noninvasive extracellular pH (pHe ) mapping with Biosensor Imaging of Redundant Deviation in Shifts (BIRDS) using MR spectroscopic imaging (MRSI) has been demonstrated on 3T clinical MR scanners at 8 mm3 spatial resolution and applied to study various liver cancer treatments. Although pHe imaging at higher resolution can be achieved by extending the acquisition time, a postprocessing method to increase the resolution is preferable, to minimize the duration spent by the subject in the MR scanner. In this work, we propose to improve the spatial resolution of pHe mapping with BIRDS by incorporating anatomical information in the form of multiparametric MRI and using an unsupervised deep-learning technique, Deep Image Prior (DIP). Specifically, we used high-resolution , , and diffusion-weighted imaging (DWI) MR images of rabbits with VX2 liver tumors as inputs to a U-Net architecture to provide anatomical information. U-Net parameters were optimized to minimize the difference between the output super-resolution image and the experimentally acquired low-resolution pHe image using the mean-absolute error. In this way, the super-resolution pHe image would be consistent with both anatomical MR images and the low-resolution pHe measurement from the scanner. The method was developed based on data from 49 rabbits implanted with VX2 liver tumors. For evaluation, we also acquired high-resolution pHe images from two rabbits, which were used as ground truth. The results indicate a good match between the spatial characteristics of the super-resolution images and the high-resolution ground truth, supported by the low pixelwise absolute error.

Predicting the pair correlation functions of silicate and borosilicate glasses using machine learning.

Glasses offer a broad range of tunable thermophysical properties that are linked to their compositions. However, it is challenging to establish a universal composition-property relationship of glasses due to their enormous compositions and chemical space. Here, we address this problem and develop a metamodel of the composition-atomistic structure relationship of a class of glassy materials via a machine learning (ML) approach. Within this ML framework, an unsupervised deep learning technique, viz., a convolutional neural network (CNN) autoencoder, and a regression algorithm, viz. random forest (RF), are integrated into a fully automated pipeline to predict the spatial distribution of atoms in a glass. The RF regression model predicts the pair correlation function of a glass in a latent space. Subsequently, the decoder of the CNN converts the latent space representation to the actual pair correlation function of the given glass. The atomistic structures of silicate (SiO2) and sodium borosilicate (NBS) based glasses with varying compositions and dopants are collected from molecular dynamics (MD) simulations to establish and validate this ML pipeline. The model is found to predict the atom pair correlation functions for many unknown glasses very accurately. This method is very generic and can accelerate the design, discovery, and fundamental understanding of the composition-atomistic structure relationship of glasses and other materials.

Hierarchical cumulative network for unsupervised medical image registration

Unsupervised deep learning techniques have gained increasing popularity in deformable medical image registration However, existing methods usually overlook the optimal similarity position between moving and fixed images To tackle this issue, we propose a novel hierarchical cumulative network (HCN), which explicitly considers the optimal similarity position with an effective Bidirectional Asymmetric Registration Module (BARM). The BARM simultaneously learns two asymmetric displacement vector fields (DVFs) to optimally warp both moving images and fixed images to their optimal similar shape along the geodesic path. Furthermore, we incorporate the BARM into a Laplacian pyramid network with hierarchical recursion, in which the moving image at the lowest level of the pyramid is warped successively for aligning to the fixed image at the lowest level of the pyramid to capture multiple DVFs. We then accumulate these DVFs and up-sample them to warp the moving images at higher levels of the pyramid to align to the fixed image of the top level. The entire system is end-to-end and jointly trained in an unsupervised manner. Extensive experiments were conducted on two public 3D Brain MRI datasets to demonstrate that our HCN outperforms both the traditional and state-of-the-art registration methods. To further evaluate the performance of our HCN, we tested it on the validation set of the MICCAI Learn2Reg 2021 challenge. Additionally, a cross-dataset evaluation was conducted to assess the generalization of our HCN. Experimental results showed that our HCN is an effective deformable registration method and achieves excellent generalization performance.

Unsupervised deep learning techniques for automatic detection of plant diseases: reducing the need of manual labelling of plant images

Crop protection from diseases through applications of plant protection products is crucial to secure worldwide food production. Nevertheless, sustainable management of plant diseases is an open challenge with a major role in the economic and environmental impact of agricultural activities. A primary contribution is expected to come from precision crop protection approaches, with treatments tailored to spatial and time-specific needs of the crop, in contrast to the current practice of applying treatments uniformly to fields. In view of this, image-based automatic detection of early disease symptoms is considered a key enabling technology for high throughput scouting of the crop, in order to timely target the treatments on emerging infection spots. Thanks to the unprecedented performance in image-recognition problems, Deep Learning (DL) methods based on Convolutional Neural Networks (CNNs) have recently entered the domain of plant disease detection. This work develops two DL approaches for automatic recognition of powdery mildew disease on cucumber leaves, with a specific focus on exploring unsupervised techniques to overcome the need of large training set of manually labelled images. To this aim, autoencoder networks were implemented for unsupervised detection of disease symptoms through: i) clusterization of features in a compressed space; ii) anomaly detection. The two proposed approaches were applied to multispectral images acquired during in-vivo experiments, and the obtained results were assessed by quantitative indices. The clusterization approach showed only partially capability to provide accurate disease detection, even if it gathered some relevant information. Anomaly detection showed instead to possess a significant potential of discrimination which could be further exploited as a prior step to train more powerful supervised architectures with a very limited number of labelled samples.

Unsupervised Probabilistic Anomaly Detection Over Nominal Subsystem Events Through a Hierarchical Variational Autoencoder

This work develops a versatile approach to discover anomalies in operational data for nominal (i.e., non-parametric) subsystem event signals using unsupervised Deep Learning techniques. Firstly, it builds a neural convolutional frameworkto extract both intrasubsystem and intersubsystem patterns. This is done by applying banks of voxel filters on the charteddata. Secondly, it generalizes the learned embedded regularity of a Variational Autoencoder manifold by merging latentspace-overlapping deviations with non-overlapping synthetic irregularities. Contingencies like novel data, modeldrift, etc., are therefore seamlessly managed by the proposed data-augmented approach. Finally, it creates a smooth diagnosis probabilistic function on the ensuing low-dimensional distributed representation. The resulting enhanced solution warrants analytically strong tools for a critical industrial environment. It also facilitates its hierarchical integrability, and provides visually interpretable insights of the degraded condition hazard to increase the confidence in its predictions. This strategy has been validated with eight pairwise-interrelated subsystems from high-speed trains. Its outcome also leads to further reliable explainability from a causal perspective.

Shadow removal method of soil surface image based on GAN used for estimation of farmland soil moisture content

It is important to obtain soil moisture content (SMC) in farmland, and soil surface images can be used to rapidly estimate SMC. The objective of this study was to propose a shadow removal algorithm to eliminate the effect of shadows in soil surface images, so as to improve the accuracy of SMC estimation. The structure of the proposed soil shadow generative adversarial networks (SS GAN) was a circulating network, which is an unsupervised method and does not require paired shadow image sets for network training. Four loss functions were defined for the network to effectively remove shadows and ensure texture detail and color consistency. This method is compared with traditional methods, supervised and unsupervised deep learning techniques by comparative experiments. Evaluations were made from visual and quantitative comparisons. Visually, the best shadow removal method was proved, it almost has no shadow boundaries or shadow areas visible for samples. The peak signal to noise ratio (PSNR) and structural similarity (SSIM) were used to quantitatively compare shadow removal images with real non-shadow images. The PSNR and SSIM of SS GAN were 28.46 and 0.95 respectively, which are superior to other methods, indicating that the images processed by SS GAN were closer to the real non-shadow images. Field experiments results shown that SS GAN has excellent shadow removal performance in the self-developed vehicle-mounted detection system. In order to verify the improvement effect of shadow removal image on SMC estimation accuracy, further field test was conducted to estimate SMC. Compared with SMC estimation results before and after shadow removal, R 2 increased from 0.69 to 0.76, and root mean square error decreased from 1.39 to 0.94%. The results show that the proposed method can effectively remove the shadow of soil image and improve the accuracy of SMC estimation in farmland.

AN UNSUPERVISED MALWARE DETECTION SYSTEM FOR WINDOWS BASED SYSTEM CALL SEQUENCES

Malware attacks have grown in prominence in recent years, posing severe security risks and resulting in significant financial losses. The ability to rapidly and reliably classify malware is vital to cybersecurity due to the exponential growth of malware variants. The role of artificial intelligence plays a significant role in cybersecurity industry. Recently, in the field of malware detection deep learning technique seeks more attention than the machine learning techniques due to the complexity of its behavior. Because the deep learning technique performs well than the machine learning techniques in terms of accuracy and it is well suited for large amount of data. The input attribute for the proposed model is windows-based system call sequence which is collected from NT mal detect project. In this work, the unsupervised deep learning technique used for text classification namely LSTM autoencoder and the performance of proposed model compares with existing DL methods such as CNN, RNN and LSTM with the performance parameters of accuracy, precision, recall and F1-measure.

A Denoising Method Using Deep Image Prior to Human-Target Detection Using MIMO FMCW Radar

A Multiple-Input Multiple-Output (MIMO) Frequency-Modulated Continuous Wave (FMCW) radar can provide a range-angle map that expresses the signal power against each range and angle. It is possible to estimate object locations by detecting the signal power that exceeds a threshold using an algorithm, such as Constant False Alarm Rate (CFAR). However, noise and multipath components often exist over the range-angle map, which could produce false alarms for an undesired location depending on the threshold setting. In other words, the threshold setting is sensitive in noisy range-angle maps. Therefore, if the noise is reduced, the threshold can be easily set to reduce the number of false alarms. In this paper, we propose a method that improves the CFAR threshold tolerance by denoising a range-angle map using Deep Image Prior (DIP). DIP is an unsupervised deep-learning technique that enables image denoising. In the proposed method, DIP is applied to the range-angle map calculated by the Curve-Length (CL) method, and then the object location is detected over the denoised range-angle map based on Cell-Averaging CFAR (CA-CFAR), which is a typical threshold setting algorithm. Through the experiments to estimate human locations in indoor environments, we confirmed that the proposed method with DIP reduced the number of false alarms and estimated the human location accurately while improving the tolerance of the threshold setting, compared to the method without DIP.

Anomaly detection on household appliances based on variational autoencoders

Electrical anomalies in residential buildings represent a serious problem that can unpredictably change the power profiles of end-users, causing a sub-optimal energy distribution. In addition, electrical faults can cause unnoticed energy wastages and higher energy bills, or even severe damages for properties and people in the most critical cases. In this paper, we introduce a novel anomaly detection method for detecting electrical faults in household appliances based on the analysis of their power signatures with unsupervised deep learning techniques. For this purpose, we trained a variational autoencoder to reconstruct the power signatures of three commonly used devices: the dishwasher, the washing machine and the dryer. For each use case, we injected several randomly generated anomalies that simulate to our best realistic electrical faults in these devices. To demonstrate the effectiveness of our method, we compared the accuracy of the variational autoencoder with the classification performance of a one-class support vector machine (OC-SVM) trained with two manual features: the energy consumption and duration of the appliance’s operations. The variational autoencoder showed higher classification accuracy with respect to the OC-SVM, reporting an F1-score greater than 90% in all the use cases. Most importantly, the results demonstrate that deep anomaly detection methods outperform traditional algorithms based on handcrafted features, allowing to better characterize the set of normal cycles and produce more precise alerts for the monitored devices.

Deep generative modeling for volume reconstruction in cryo-electron microscopy

Advances in cryo-electron microscopy (cryo-EM) for high-resolution imaging of biomolecules in solution have provided new challenges and opportunities for algorithm development for 3D reconstruction. Next-generation volume reconstruction algorithms that combine generative modelling with end-to-end unsupervised deep learning techniques have shown promise, but many technical and theoretical hurdles remain, especially when applied to experimental cryo-EM images. In light of the proliferation of such methods, we propose here a critical review of recent advances in the field of deep generative modelling for cryo-EM reconstruction. The present review aims to (i) provide a unified statistical framework using terminology familiar to machine learning researchers with no specific background in cryo-EM, (ii) review the current methods in this framework, and (iii) outline outstanding bottlenecks and avenues for improvements in the field.

Quartet Based Gene Tree Imputation Using Deep Learning Improves Phylogenomic Analyses Despite Missing Data.

Species tree estimation is frequently based on phylogenomic approaches that use multiple genes from throughout the genome. However, for a combination of reasons (ranging from sampling biases to more biological causes, as in gene birth and loss), gene trees are often incomplete, meaning that not all species of interest have a common set of genes. Incomplete gene trees can potentially impact the accuracy of phylogenomic inference. We, for the first time, introduce the problem of imputing the quartet distribution induced by a set of incomplete gene trees, which involves adding the missing quartets back to the quartet distribution. We present Quartet based Gene tree Imputation using Deep Learning (QT-GILD), an automated and specially tailored unsupervised deep learning technique, accompanied by cues from natural language processing, which learns the quartet distribution in a given set of incomplete gene trees and generates a complete set of quartets accordingly. QT-GILD is a general-purpose technique needing no explicit modeling of the subject system or reasons for missing data or gene tree heterogeneity. Experimental studies on a collection of simulated and empirical datasets suggest that QT-GILD can effectively impute the quartet distribution, which results in a dramatic improvement in the species tree accuracy. Remarkably, QT-GILD not only imputes the missing quartets but can also account for gene tree estimation error. Therefore, QT-GILD advances the state-of-the-art in species tree estimation from gene trees in the face of missing data.

Towards unsupervised classification of macromolecular complexes in cryo electron tomography: Challenges and opportunities

Background and Objectives: Cryo electron tomography visualizes native cells at nanometer resolution, but analysis is challenged by noise and artifacts. Recently, supervised deep learning methods have been applied to decipher the 3D spatial distribution of macromolecules. However, in order to discover unknown objects, unsupervised classification techniques are necessary. In this paper, we provide an overview of unsupervised deep learning techniques, discuss the challenges to analyze cryo-ET data, and provide a proof-of-concept on real data.Methods: We propose a weakly supervised subtomogram classification method based on transfer learning. We use a deep neural network to learn a clustering friendly representation able to capture 3D shapes in the presence of noise and artifacts. This representation is learned here from a synthetic data set.Results: We show that when applying k-means clustering given a learning-based representation, it becomes possible to satisfyingly classify real subtomograms according to structural similarity. It is worth noting that no manual annotation is used for performing classification.Conclusions: We describe the advantages and limitations of our proof-of-concept and raise several perspectives for improving classification performance.

Implementing Supervised and Unsupervised Deep-Learning Methods to Predict Sputtering Plasma Features, a Step toward Digitizing Sputter Deposition of Thin Films

The spectral emission data from the plasma glow of various sputtering targets containing indium oxide, zinc oxide, and tin oxide were obtained. The plasma was generated at various power and chamber pressures. These spectral data were then converted into two-dimensional arrays by implementing a basic array-reshaping technique and a more complex procedure utilizing an unsupervised deep-learning technique, known as the self-organizing-maps method. The two-dimensional images obtained from each single-emission spectrum of the plasma mimic an image that can then be used to train a convolutional neural network model capable of predicting certain plasma features, such as impurity levels in the sputtering target, working gas composition, plasma power, and chamber pressure during the machine operation. We show that our single-array-to-2D-array conversion technique, coupled with deep-learning techniques and computer vision, can achieve high predictive accuracy and can, therefore, be fundamental to the construction of a sputtering system’s digital twin.

Spatio-temporal contrastive learning for acoustic data

Large volumes of water column sonar data have been generated from acoustic surveys of living marine resources. They provide valuable information about marine ecosystems. To leverage them for acoustic target identification, scarcity of annotations is usually an issue. While some annotations have been generated by scientists via manual scrutiny or limited automation, they are limited to certain species. To fill in this gap, we propose a spatio-temporal contrastive learning approach for acoustic data. This unsupervised deep learning technique leverages both acoustic and non-acoustic (spatial and temporal) information for acoustic target classification. We firstly employ the Simple Linear Iterative Clustering (SLIC) algorithm to extract superpixels from acoustic data. Then the spatial and temporal similarity between two superpixels is computed, and the proposed spatio-temporal contrastive learning approach is applied to learn a semantically meaningful representation for superpixels. Finally, the learned representations of superpixels are used in acoustic target classification. We demonstrate that this approach outperforms previous methods with acoustic data collected by the NOAA National Marine Fisheries Service in the Gulf of Maine and Georges Bank region over 10 years. This dataset is archived at NOAA National Centers for Environmental Information and accessed for free through AWS and NOAA Big Data Program.

Deep‐learning‐based line enhancer for passive sonar systems

Detection of acoustic tonals from surfaces and underwater vehicles is important for passive sonar systems. Enhancements of the tonals are usually necessary in passive sonar prior to detections. Conventionally, passive sonars employ adaptive line enhancers (ALE) in order to realise enhancements of the tonals. However, ALEs have requirements on their input signal-to-noise ratios (SNR). When the SNR inputs are too low, the ALEs cannot perform well. Therefore, for the purpose of overcoming the limitations of the SNR inputs to ALEs, this study proposes to enhance tonals using unsupervised deep-learning techniques. The proposed deep-learning-based line enhancer (DLE) is based on an autoencoder neural network. The simulation results show that when the input SNR is −28 dB, the proposed DLE still achieves an SNR gain of 15 dB. However, the reference ALE fails. The experimental results also demonstrate the superiority of the proposed DLE.

Combining Supervised and Unsupervised Learning Algorithms for Human Activity Recognition.

Human activity recognition is an extensively researched topic in the last decade. Recent methods employ supervised and unsupervised deep learning techniques in which spatial and temporal dependency is modeled. This paper proposes a novel approach for human activity recognition using skeleton data. The method combines supervised and unsupervised learning algorithms in order to provide qualitative results and performance in real time. The proposed method involves a two-stage framework: the first stage applies an unsupervised clustering technique to group up activities based on their similarity, while the second stage classifies data assigned to each group using graph convolutional networks. Different clustering techniques and data augmentation strategies are explored for improving the training process. The results were compared against the state of the art methods and the proposed model achieved 90.22% Top-1 accuracy performance for NTU-RGB+D dataset (the performance was increased by approximately 9% compared with the baseline graph convolutional method). Moreover, inference time and total number of parameters stay within the same magnitude order. Extending the initial set of activities with additional classes is fast and robust, since there is no required retraining of the entire architecture but only to retrain the cluster to which the activity is assigned.

A Tour of Unsupervised Deep Learning for Medical Image Analysis.

Interpretation of medical images for the diagnosis and treatment of complex diseases from high-dimensional and heterogeneous data remains a key challenge in transforming healthcare. In the last few years, both supervised and unsupervised deep learning achieved promising results in the area of medical image analysis. Several reviews on supervised deep learning are published, but hardly any rigorous review on unsupervised deep learning for medical image analysis is available. The objective of this review is to systematically present various unsupervised deep learning models, tools, and benchmark datasets applied to medical image analysis. Some of the discussed models are autoencoders and their variants, Restricted Boltzmann Machines (RBM), Deep Belief Networks (DBN), Deep Boltzmann Machine (DBM), and Generative Adversarial Network (GAN). Future research opportunities and challenges of unsupervised deep learning techniques for medical image analysis are also discussed. Currently, interpretation of medical images for diagnostic purposes is usually performed by human experts that may be replaced by computer-aided diagnosis due to advancement in machine learning techniques, including deep learning, and the availability of cheap computing infrastructure through cloud computing. Both supervised and unsupervised machine learning approaches are widely applied in medical image analysis, each of them having certain pros and cons. Since human supervisions are not always available or are inadequate or biased, therefore, unsupervised learning algorithms give a big hope with lots of advantages for biomedical image analysis.

IoT Botnet Anomaly Detection Using Unsupervised Deep Learning

The Internet of Things has become a cutting-edge technology that is continuously evolving in size, connectivity, and applicability. This ecosystem makes its presence felt in every aspect of our lives, along with all other emerging technologies. Unfortunately, despite the significant benefits brought by the IoT, the increased attack surface built upon it has become more critical than ever. Devices have limited resources and are not typically created with security features. Lately, a trend of botnet threats transitioning to the IoT environment has been observed, and an army of infected IoT devices can expand quickly and be used for effective attacks. Therefore, identifying proper solutions for securing IoT systems is currently an important and challenging research topic. Machine learning-based approaches are a promising alternative, allowing the identification of abnormal behaviors and the detection of attacks. This paper proposes an anomaly-based detection solution that uses unsupervised deep learning techniques to identify IoT botnet activities. An empirical evaluation of the proposed method is conducted on both balanced and unbalanced datasets to assess its threat detection capability. False-positive rate reduction and its impact on the detection system are also analyzed. Furthermore, a comparison with other unsupervised learning approaches is included. The experimental results reveal the performance of the proposed detection method.

Deep learning for whole-body medical image generation.

Artificial intelligence (AI) algorithms based on deep convolutional networks have demonstrated remarkable success for image transformation tasks. State-of-the-art results have been achieved by generative adversarial networks (GANs) and training approaches which do not require paired data. Recently, these techniques have been applied in the medical field for cross-domain image translation. This study investigated deep learning transformation in medical imaging. It was motivated to identify generalizable methods which would satisfy the simultaneous requirements of quality and anatomical accuracy across the entire human body. Specifically, whole-body MR patient data acquired on a PET/MR system were used to generate synthetic CT image volumes. The capacity of these synthetic CT data for use in PET attenuation correction (AC) was evaluated and compared to current MR-based attenuation correction (MR-AC) methods, which typically use multiphase Dixon sequences to segment various tissue types. This work aimed to investigate the technical performance of a GAN system for general MR-to-CT volumetric transformation and to evaluate the performance of the generated images for PET AC. A dataset comprising matched, same-day PET/MR and PET/CT patient scans was used for validation. A combination of training techniques was used to produce synthetic images which were of high-quality and anatomically accurate. Higher correlation was found between the values of mu maps calculated directly from CT data and those derived from the synthetic CT images than those from the default segmented Dixon approach. Over the entire body, the total amounts of reconstructed PET activities were similar between the two MR-AC methods, but the synthetic CT method yielded higher accuracy for quantifying the tracer uptake in specific regions. The findings reported here demonstrate the feasibility of this technique and its potential to improve certain aspects of attenuation correction for PET/MR systems. Moreover, this work may have larger implications for establishing generalized methods for inter-modality, whole-body transformation in medical imaging. Unsupervised deep learning techniques can produce high-quality synthetic images, but additional constraints may be needed to maintain medical integrity in the generated data.

Accelerating Ecological Sciences from Above: Spatial Contrastive Learning for Remote Sensing

The rise of neural networks has opened the door for automatic analysis of remote sensing data. A challenge to using this machinery for computational sustainability is the necessity of massive labeled data sets, which can be cost-prohibitive for many non-profit organizations. The primary motivation for this work is one such problem; the efficient management of invasive species -- invading flora and fauna that are estimated to cause damages in the billions of dollars annually. As an ongoing collaboration with the New York Natural Heritage Program, we consider the use of unsupervised deep learning techniques for dimensionality reduction of remote sensing images, which can reduce sample complexity for downstream tasks and decreases the need for large labeled data sets. We consider spatially augmenting contrastive learning by training neural networks to correctly classify two nearby patches of a landscape as such. We demonstrate that this approach improves upon previous methods and naive classification for a large-scale data set of remote sensing images derived from invasive species observations obtained over 30 years. Additionally, we simulate deployment in the field via active learning and evaluate this method on another important challenge in computational sustainability -- landcover classification -- and again find that it outperforms previous baselines.