Decoding regulatory structures and features from epigenomics profiles: A Roadmap-ENCODE Variational Auto-Encoder (RE-VAE) model.

<p><strong>Deep generative models have greatly advanced the field of artificial intelligence by learning the distribution of unlabelled datasets. In this thesis, we aim to develop a novel variational autoencoder (VAE)-based deep generative model that learns meaningful representations for effective unsupervised clustering in the latent space of high-dimensional datasets.</strong></p><p>Existing VAE-based deep clustering algorithms assume a mixture distribution either for the variational posterior over the latent variables or for the prior, to approximate the distribution of unlabelled datasets. The proposed mixtures variational autoencoders (MVAE) model assumes a mixture distribution for both the prior and variational posterior over the latent variables within the VAE framework.</p><p>We apply the variational inference to construct the evidence lower bound (ELBO) of the marginal log-likelihood function for our proposed MVAE model and then derive the analytical form for the clustering assignment estimate, also called the posterior probability estimate. This estimate is associated with each component of the VAE framework and enables the optimization of two distinct ELBO functions based on soft and hard assignment approaches. The training procedure for MVAE involves the joint optimization of both clustering assignment probabilities and model parameters. We evaluate the clustering performance on the latent embeddings of a dataset using both the posterior probability estimate and the Gaussian mixture model (GMM) methods.</p><p>We proposed two variants of the expectation-maximization (EM) algorithm to optimize the parameters for the MVAE model. The MVAE (EM) algorithm provides two distinct, numerically stable solutions for the posterior probability estimate: MVAE(EM-V1), where the estimate is associated with only the prior distribution, and MVAE(EM-V2), where the estimate incorporates both the variational posterior and the prior distribution. Among the proposed models, MVAE(EM-V2) requires less computation time for training. Additionally, this model achieves superior clustering performance on most datasets compared to the baseline models.</p><p>We modified the MVAE(EM-V2) algorithm by including a coefficient (β) with the regularizer term. The trained regularized mixtures VAE model with a small regularization coefficient (β < 1) achieves good unsupervised clustering performance on the test datasets. We further propose a variant of the regularized mixtures VAE model in which the regularization coefficient follows an annealing schedule from β > 1 to β < 1. The scheduled regularized model demonstrates superior clustering accuracy across most benchmark datasets compared to state-of-the-art deep clustering algorithms.</p><p>In this thesis, we point out that assuming a mixture distribution for both the prior and variational posterior components over the latent variables within the VAE framework enhances unsupervised clustering in the latent space. Our proposed models outperform state-of-the-art deep clustering algorithms, including VADE and k-DVAE, as well as the standard VAE model, in cluster analysis of latent representations across most benchmark datasets. Additionally, the proposed models demonstrate reasonable reconstruction performance and generate realistic examples from the latent space.</p>

This paper evaluates the effectiveness of advanced Variational Autoencoder (VAE) models in overcoming latent space entanglement and insufficient disentanglement, common issues in traditional VAEs. Traditional VAEs often face challenges in separating distinct features within the latent space, leading to entangled representations that hinder interpretability and compression efficiency. The advanced VAE models examined in this study address these issues by enhancing disentanglement, which results in clearer separation of latent factors and more interpretable representations. However, this improvement in disentanglement may result in a trade-off with reconstruction quality. The article shows that, while these sophisticated models improve disentanglement, they may also have worse reconstruction quality than classic VAEs. The findings highlight the necessity of hyperparameter optimization in successfully navigating this trade-off. Future study should investigate novel model architectures and hyperparameter optimization strategies to optimize the balance of disentanglement and reconstruction quality. Overall, the research emphasizes the ability of advanced VAE models to generate more interpretable representations and the importance of careful tuning to resolve the inherent trade-offs.

A variational autoencoder (VAE) is a generational deep learning model based on encoding a particular observation into a latent space and then decoding it while incorporating some random noise with the intuition of being able to generate slightly different forms of the input observation. Here, I propose a VAE that is able to generate denovo drug compounds by feeding in a known set of drug compounds against a particular biological target. To demonstrate the ability of the proposed VAE as a generative model, known drug molecules belonging to the class of Neuraminidase (NA) inhibitors were taken from the ZINC database and fed into the VAE model as one-hot encoded SMILES strings. Similarly, active NA inhibitors and decoy molecules together were also fed to compare efficiency. The generated molecules were then screened to remove impractical structures. Next, a drug-likeness (QED) score was computed for each candidate molecule and a cutoff of 0.5 was used to extract viable candidates. To ensure that the generated drug compounds were active NA inhibitors, a series of Artificial Neural Networks (ANNs) classifiers based on three different characterization techniques, namely chemical fingerprinting, molecular descriptions and graph convolutions, were developed to identify active NA inhibitors from decoy molecules. The feature candidate data were then fed into the three designed ANNs to obtain the final set of novel, viable and active NA inhibitors. Seventy-one new NA inhibitors were obtained after three runs of the VAE model under different parameterizations. The proposed VAE can hence be used to generate de-novo drug compounds for a wide variety of biological targets.

BackgroundThe early diagnosis and monitoring of Alzheimer's disease (AD) presents a significant challenge due to its heterogeneous nature, which includes variability in cognitive symptoms, diagnostic test results, and progression rates. This study aims to enhance the understanding of AD progression by integrating neuroimaging metrics with demographic data using a novel machine learning technique.MethodWe used supervised Variational Autoencoders (VAEs), a generative AI method, to analyze high‐dimensional neuroimaging data for AD progression, incorporating age and gender as covariates. We used two non‐overlapping datasets: 257 samples from ADNI3 and 676 from ADNI2, analyzing 68 cortical and 48 subcortical grey matter volumes extracted from their MRI 3D T1 images. The VAE model aimed to minimize reconstruction error (MSE), Kullback‐Leibler divergence, and classification error, and to estimate latent variables for each subject. A Disease Progression Score (DPS) was calculated by encoding the extracted imaging features into a latent space and projecting test data onto a trajectory curve. The study used stratified sampling for robustness and assessed the model's performance using the area under the ROC curve (AUC), and correlated the mean DPS with cognitive assessment scores by applying Kendall’s Tau.ResultThe VAE model demonstrated excellent discriminative power in classifying AD progression stages, with ROC AUC values near 1, particularly when using all 116 features. Cortical volumes were more predictive than subcortical volumes. The mean DPS showed a statistically significant correlation with cognitive assessments (p<0.01), with Kendall's Tau values of 0.66 for CDR‐SB, ‐0.45 for MMSE, and ‐0.49 for MoCA, indicating its validity as a quantitative biomarker for cognitive decline in AD.ConclusionSupervised VAEs can effectively model the progression of Alzheimer’s disease brain atrophy on MRI and could potentially serve as imaging biomarkers in clinical trials. The strong correlation between the DPS and cognitive assessments highlights the potential of supervised VAEs to provide a quantifiable measure of AD severity, which would be useful for clinical assessments and for objectively monitoring disease progression. The study's methodology and findings contribute to computational neuroscience and offer a foundation for future research in early detection and personalized treatment strategies for Alzheimer's disease.

Recognizing human actions accurately from video data is a difficult task because of the variability in human motion, appearance, and context. A new approach to human action recognition is proposed in this study, which uses a combination of Variational Autoencoder (VAE) and Transformer models. The VAE with IWGECO optimization is utilized for acquiring a compact and discriminative representation of the input 2D skeletons from video frames along with its spatial relations. This encoded representation of 2D poses is then fed into a Transformer model for sequence modeling and classification, resulting in effective recognition of human actions. The proposed approach is evaluated on the MPOSE2021 benchmark dataset, outperforming existing methods (90.6%) with an average classification accuracy of 92.0%. The model performs well on real-life data too showcasing its generalizing potential. The study also addresses the domain shift problem occurring in real-world situations by using a pre-trained VAE and subsequently training the Transformer on the MPOSE2021 dataset. The results demonstrate the effectiveness of combining VAE and Transformer models for human action recognition.

The accurate land cover change detection is critical to improve the landscape dynamics analysis and mitigate desertification problems efficiently. Desertification detection is a challenging problem because of the high degree of similarity between some desertification cases and like-desertification phenomena, such as deforestation. This article provides an effective approach to detect deserted regions based on Landsat imagery and variational autoencoder (VAE). The VAE model, as a deep learning-based model, has gained special attention in features extraction and modeling due to its distribution-free assumptions and superior nonlinear approximation. Here, a VAE approach is applied to spectral signatures for detecting pixels affected by the land cover change. The considered features are extracted from multitemporal images and include multispectral information, and no prior image segmentation is required. The proposed method was evaluated on the publicly available remote sensing data using multitemporal Landsat optical images taken from the freely available Landsat program. The arid region around Biskra in Algeria is selected as a study area since it is well-known that desertification phenomena strongly influence this region. The VAE model was evaluated and compared with restricted Boltzmann machines, deep learning model, and binary clustering algorithms, including Agglomerative, BIRCH, expected maximization, k-mean clustering algorithms, and one-class support vector machine. The comparative results showed that the VAE consistently outperformed the other models for detecting changes to the land cover, mainly deserted regions. This study also showed that VAE outperformed the state-of-the-art algorithms.

ABSTRACTThe immune defense function protecting the body from invasive pathogens is a key indicator of an individual's health and lacks of methods for quantitative evaluation. This study introduces ImmuDef, a novel algorithm for precisely and quantitatively assessing anti‐infection immune defense function based on RNA‐seq data. ImmuDef selects immune signatures through comparisons of acquired immunodeficiency syndrome (AIDS) or severe sepsis vs. healthy controls (HC) and reduces dimension to construct a latent space via a variational autoencoder (VAE) model (QImmuDef‐VAE), a representation deep learning model. Based on this model, a defense immune score (DImmuScore) was calculated by measuring the distance between a patient and HC within latent space. We validated ImmuDef on 3202 samples across four immune states: immunodeficiency, immunocompromised, immunocompetent, and immunoactive. As a result, DImmuScore achieves high classification accuracy (mean accuracy: 71.75%–76.25%) among samples with various immune states and infections. Furthermore, DImmuScore can serve as a metric for infectious disease severity, where its gradient directly quantifies disease severity. As an application, DImmuScore can be a strong prognostic indicator, effectively stratifying mortality/survival in both sepsis and COVID‐19 patients with no symptomatic difference. This framework was validated across five infectious diseases, establishing the first quantitative standard for cross‐disease immune defense assessment.

Biodiversity monitoring in forests requires scalable, automated tools for detecting ecological anomalies across time and space.This paper reports on a three-month pilot deployment (April 1 to June 30, 2025) in Dyrehaven, an 11 km 2 forest park near Copenhagen, Denmark, where acoustic data from 10 distributed AudioMoth sensors and vegetation indices from Sentinel-2 imagery were collected.We trained separate variational autoencoder (VAE) models on each modality to test the technical feasibility of learning ecological baselines.Since no ecological anomalies occurred during the observation period, evaluation focused on reconstruction errors, which indicate how well VAEs can capture typical site-specific ecological patterns (i.e., baseline modeling).Both acoustic and satellite pipelines achieved low reconstruction errors, demonstrating that VAEs can reliably model normal ecological dynamics.This establishes the foundation for future studies on anomaly detection, which will require larger datasets containing true ecological anomalies identified and labeled by experts.Ongoing work focuses on extending data collection to additional forest sites, while future anomaly detection will require expert-labeled anomalies to calibrate baselines and validate model performance for robust, multimodal biodiversity monitoring.

We introduce a learning model able to conceal personal information (e.g. gender, age, ethnicity, etc.) from an image while maintaining any additional information present in the image (e.g. smile, hair-style, brightness). Our trained model is not provided the information that it is concealing, and does not try learning it either. Namely, we created a variational autoencoder (VAE) model that is trained on a dataset including labels of the information one would like to conceal (e.g. gender, ethnicity, age). These labels are directly added to the VAE's sampled latent vector. Due to the limited number of neurons in the latent vector and its appended noise, the VAE avoids learning any relation between the given images and the given labels, as those are given directly. Therefore, the encoded image lacks any of the information one wishes to conceal. The encoding may be decoded back into an image according to any provided properties (e.g. a 40-year old woman). Our method successfully conceals the private information; a convolutional neural network trained on the concealed images cannot restore the original private information. In contrast to the private information, a user study shows that the remaining properties of the original image carry-on to the concealed image. The proposed architecture can be used as a mean for privacy preserving and can serve as an input to systems, which will become unbiased and not suffer from prejudice.

Auto-Encoder (AE) based neural generative frameworks model the joint-distribution between the data and the latent space using an Encoder-Decoder pair, with regularization imposed in terms of a prior over the latent space. Despite their advantages, such as stability in training, efficient inference, the performance of AE based models has not reached the superior standards of the other generative models such as Generative Adversarial Networks (GANs). Motivated by this, we examine the effect of the latent prior on the generation quality of deterministic AE models in this paper. Specifically, we consider the class of Generative AE models with deterministic Encoder-Decoder pair (such as Wasserstein Auto-Encoder (WAE), Adversarial Auto-Encoder (AAE)), and show that having a fixed prior distribution, \textit{a priori}, oblivious to the dimensionality of the `true' latent space, will lead to the infeasibility of the optimization problem considered. As a remedy to the issue mentioned above, we introduce an additional state space in the form of flexibly learnable latent priors, in the optimization objective of WAE/AAE. Additionally, we employ a latent-space interpolation based smoothing scheme to address the non-smoothness that may arise from highly flexible priors. We show the efficacy of our proposed models, called FlexAE and FlexAE-SR, through several experiments on multiple datasets, and demonstrate that FlexAE-SR is the new state-of-the-art for the AE based generative models in terms of generation quality as measured by several metrics such as Fr\'echet Inception Distance, Precision/Recall score. Code for our paper is available at: \url{https://github.com/dair-iitd/FlexAE}

This paper presents a novel approach for detecting sensor failures using image-based feature representation and the Convolutional Variational Autoencoder (CVAE) model. Existing methods are limited when analyzing multiple failure modes simultaneously or adapting to diverse sensor data. This limitation may compromise decision-making and system performance, hence the need for more flexible and resilient models. The proposed approach transforms sensor data into image-based feature representations of statistics such as mean, variance, kurtosis, entropy, skewness, and correlation. The CVAE is trained on such image representations, and the corresponding reconstruction error leads to a Health Index (HI) for detecting multiple sensor failures. Moreover, the CVAE latent space is used to define a complementary HI and a convenient visualization tool, enhancing the interpretability of the proposed approach. The evaluation of the proposed detection approach with data comprising diverse configurations of faulty sensors showed encouraging results. The proposed approach is illustrated in an industrial case study emerging from the aeronautical domain, with data from a complex electromechanical system comprising nearly 80 sensor measurements at a 1 Hz sampling rate. The results demonstrate the potential of the proposed method in detecting multiple sensor failures.

Information security is crucial with the increasing data and internet usage. In data communication, Imaging data is one of the most frequently used data types for communication, especially in intelligence operations and in law enforcement use cases. Steganography is a method for hiding secrets in cover images, followed by complex image encryption and decryption methods. Previously, many Deep Learning (DL) methods have been proposed on steganography and achieved competitive results. However, there is still a need for lightweight, computationally less expensive DL models. Therefore, a computationally inexpensive, lightweight dual-branch based Variational Auto Encoder (VAE) model is proposed, namely SE-Attn StegaVAE. In this model, the Squeeze and Excitation (SE) block, Attention, and skipping connections are effectively used, whereas a sequentially optimal way is adapted to hide the secret in attention-oriented cover samples. Furthermore, a multi-objective loss function is proposed to penalize the model to hide secrets effectively and reconstruct them without any loss of information. In this study, three experiments have been performed with three different bit embeddings (2, 4, and 8) on the DIV2K dataset utilizing the SE-Attn StegaVAE model. Two-fold validation-based testing results outperformed as compared to State-of-the-Art (SOTA) methods and proven to be more error-free, computationally less expensive, with competitive similarity scores.

Objective: False lumen (FL) thrombosis plays a key role in aortic remodeling after Frozen Elephant Trunk (FET) surgery, yet current imaging assessments are limited to categorical classifications. This study aimed to evaluate an unsupervised artificial intelligence (AI) algorithm based on a variational autoencoder (VAE) for automated, continuous quantification of FL thrombosis using serial computed tomography angiography (CTA). Methods: In this retrospective study, a VAE model was applied to axial CTA slices from 30 patients with aortic dissection who underwent FET surgery. The model encoded each image into a structured latent space, from which a continuous “thrombus score” was developed and derived to quantify the extent of FL thrombosis. Thrombus scores were compared between postoperative and follow-up scans to assess individual remodeling trajectories. Results: The VAE successfully encoded anatomical features of the false lumen into a structured latent space, enabling unsupervised classification of thrombus states. A continuous thrombus score was derived from this space, allowing slice-by-slice quantification of thrombus burden across the aorta. The algorithm demonstrated robust reconstruction accuracy and consistent separation of fully patent, partially thrombosed, and completely thrombosed lumen states without the need for manual annotation. Across the cohort, 50% of patients demonstrated an increase in thrombus score over time, 40% a decrease, and 10% remained unchanged. Despite these individual differences, no statistically significant change in overall thrombus burden was observed at the group level (p = 0.82), emphasizing the importance of individualized longitudinal assessment. Conclusions: The VAE-based method enables reproducible, annotation-free quantification of FL thrombosis and captures patient-specific remodeling patterns. This approach may enhance post-FET surveillance and supports the integration of AI-driven tools into personalized aortic imaging workflows.

In recent years, great strides have been made in the field of affective\ncomputing. Several models have been developed to represent and quantify\nemotions. Two popular ones include (i) categorical models which represent\nemotions as discrete labels, and (ii) dimensional models which represent\nemotions in a Valence-Arousal (VA) circumplex domain. However, there is no\nstandard for annotation mapping between the two labelling methods. We build a\nnovel algorithm for mapping categorical and dimensional model labels using\nannotation transfer across affective facial image datasets. Further, we utilize\nthe transferred annotations to learn rich and interpretable data\nrepresentations using a variational autoencoder (VAE). We present "LeVAsa", a\nVAE model that learns implicit structure by aligning the latent space with the\nVA space. We evaluate the efficacy of LeVAsa by comparing performance with the\nVanilla VAE using quantitative and qualitative analysis on two benchmark\naffective image datasets. Our results reveal that LeVAsa achieves high\nlatent-circumplex alignment which leads to improved downstream categorical\nemotion prediction. The work also demonstrates the trade-off between degree of\nalignment and quality of reconstructions.\n

Acoustic metamaterials are engineered microstructures with special mechanical and acoustic properties enabling exotic effects such as wave steering, focusing and cloaking. In this research, we develop a new machine learning framework for predicting optimal metastructures such as planar configurations of scatterers with specific functionalities. Specifically, we implement this framework by combining probabilistic generative modeling with deep learning and propose two models: a conditional variational autoencoder (CVAE) and a supervised variational autoencoder (SVAE) model. As an application of the method, here we design an acoustic cloak considering a minimization of total scattering cross-section (TSCS) for a set of cylindrical obstacles. We work with the sets of cylindrical objects confined in a region of space and streamline the design of configurations with minimal TSCS, demonstrating broadband cloaking effect at discrete sets of wavenumbers. After establishing the artificial neural networks that are capable of learning the TSCS based on the location of cylinders, we discuss our inverse design algorithms, combining variational autoencoders and the Gaussian process, for predicting optimal arrangements of scatterers given the TSCS. We show results for up to eight cylinders and discuss the efficiency and other advantages of the machine learning approach.