27 papers · ranked by Valyu relevance
Ata Akbari Asanjan, Milad Memarzadeh, Bryan Matthews, Nikunj C. Oza
In this study, we focus on the training process and inference improvements of deep neural networks (DNNs), specifically Autoencoders (AEs) and Variational Autoencoders (VAEs), using Random Fourier Transformation (RFT). We further explore the role of RFT in model training behavior using Frequency Principle (F-Principle)…
Evan Gorstein, Mengze Tang, Hailey Bruzzone, Claudia Solís-Lemus
Standard methods for ancestral sequence reconstruction (ASR) rely on substitution models for the residues in a biological sequence and assume independent evolution across these sites, ignoring the epistatic interactions that shape molecular evolution. In contrast, deep learning models like variational autoencoders…
Gadirov, Hamid
M AC H I N E L E A R N I N G F O R S C I E N T I F I C V I S UA L I Z AT I O N : E N S E M B L E DATA A N A LY S I S
Ahmed M. Daoud, Osama M. Elkomy, Walid I. Khedr, Khalid M. Hosny
In this paper, we propose a new hybrid model, called AVE, that integrates the strengths of Autoencoder (AE) and Variational Autoencoder (VAE) to enhance outlier detection for numerous high-dimensional datasets. The proposed architecture leverages the reconstruction strengths of AE and the regularized latent space of…
Wei Xiong, Na Xiao, Ruili Wang, Dandan Peng
To address the uneven distribution of fault categories in data sets for deep learning-based fault diagnosis, we propose a fault diagnosis framework combining an improved Variational Autoencoder (Similarity-Aware VAE) with a Wavelet-Convolutional 1D-CNN. The Similarity-Aware VAE employs a novel similarity loss function…
Edgar R. Guzman, Robert D. Howe, Zhaoyang Wang, Hieu Nguyen + 2 more
The unpredictable nature of outdoor settings introduces numerous safety concerns, making hazard detection crucial for safe navigation. To address this issue, this paper proposes a sidewalk hazard detection system that combines a Variational Autoencoder (VAE) with a One-Class Support Vector Machine (OCSVM), using a…
Konstantinos Patlatzoglou, Libor Pastika, Joseph Barker, Ewa Sieliwonczyk + 10 more
Artificial intelligence-enhanced electrocardiogram (AI-ECG) models have shown outstanding performance in diagnostic and prognostic tasks, yet their black-box nature hampers clinical adoption. Meanwhile, a growing demand for explainable AI in medicine underscores the need for transparent, trustworthy decision-making.…
Gananath R
Variational Autoencoders (VAEs) belong to a family of autoencoders with probabilistic properties, making them well suited for generating data by producing a smooth and continuous latent space. Despite being introduced over a decade ago, the method continues to be widely adopted in both research and industry for diverse…
Touseef Iqbal, Ifrah Raoof, Mohannad Alkanan, Yonis Gulzar
HIPAA breaches and unauthorized access to Electronic Health Records (EHRs) have been growing more likely due to the sudden digitalization of the healthcare sector. High endurance, privacy-based security practices have never been more in demand as hospitals and other medical facilities of this type have clung to the…
J. Solomon, Rishi Leburu, Matthias Chung
We propose the Variational Sparse Paired Autoencoder (vsPAIR) to address this challenge. The architecture pairs a standard VAE encoding observations with a sparse VAE encoding quantities of interest, connected through a learned latent mapping. The variational structure enables uncertainty estimation, the paired…
Hazhir Aliahmadi, Irina Babayan, Greg van Anders
Despite their ubiquity, variational autoencoders (VAEs) inherently suffer from posterior collapse, a failure mode in which latent variables are effectively ignored. This failure arises because explicit prior imposition drives optimization toward loss landscape regions corresponding to uninformative latent…
Linna Wu, Yong Yang, Wenhao Wang, Yuanlun Xie + 3 more
Cross-domain emotion recognition based on electroencephalogram (EEG) is a challenging task, as EEG signals collected from different subjects or at different moments exhibit significant differences in distribution. How to enable deep learning model to learn the common feature space and reduce the distribution…
Kendall Reid, Erhan Guven
High-throughput gene expression measurements are biased by technical and biological confounding variables, which obscure true biological signals. A common deep-learning-based solution involves training latent-space models with adversarial regularizers to ignore information from confounding variables. These methods rely…
Sota Okuda, Atsushi Minami, Michio Aiko, Kohei Uetsuka + 3 more
Transcriptional control arises from the specific recognition of promoter DNA by transcription factors (TFs), forming the basis of cellular information processing and gene regulation. In synthetic biology, TF-promoter interactions are assembled into gene circuits to program cellular behaviors. To ensure reliable circuit…
Sebastian Magana, Wenjun Zhao, Khanh Dao Duc
Inferring continuous morphological transformations from collections of static biological snapshots is an important, yet challenging problem. In the context of cellular biology, prevailing approaches reduce 3D shape collections to static reconstructions or hand-crafted descriptors, which fail to capture smooth…
Jesús Jaime Moreno Escobar, Mauro Daniel Castillo Pérez, Erika Yolanda Aguilar del Villar, Hugo Quintana Espinosa + 4 more
This study presents a sensor-based condition monitoring framework for the diagnosis and prognosis of neurophysiological states using electroencephalographic (EEG) signals. Leveraging a comparative deep learning architecture, we evaluate a baseline Variational Autoencoder against a Transformer-based Autoencoder to model…
John Knight
Latent Factor Analysis via Dynamical Systems (LFADS) is a powerful variational autoencoder for inferring neural population dynamics from spike train data. However, LFADS suffers from pos-terior collapse, where the learned posterior collapses to the prior, eliminating meaningful latent representations. Current solutions…
Hao Lu, Onur Can Koyun, Yongxin Guo, Zhengjie Zhu + 2 more
Vector Quantization (VQ) underpins many modern generative frameworks such as VQ-VAE, VQ-GAN, and latent diffusion models. Yet, it suffers from the persistent problem of codebook collapse, where a large fraction of code vectors remains unused during training. This work provides a new theoretical explanation by…
Corrado Pancotti, Piero Fariselli, Jonas Meisner, Anders Krogh
In this paper, we demonstrate that it is feasible to train a deep generative model for dimensionality reduction with millions of features using few samples, which makes this type of generative model a more versatile alternative to standard methods for dimensionality reduction. Specifically, we hypothesize that for a…
Abdelhakim Ziani, Andras Horvath, Paolo Ballarini
Heavy-tailed distributions are prevalent in performance evaluation, network traffic, and risk modeling. This behavior poses a fundamental challenge for modern deep generative models. Standard Variational Autoencoders (VAEs) employ Gaussian decoder likelihoods and Lipschitz-constrained neural networks, a combination…
Zeyu Fu, Jiawei Fu, Chunlin Chen, Keyang Zhang + 3 more
Single-cell omics routinely profile millions of cells across the transcriptome and the epigenome. However, embeddings used for clustering, trajectory inference, and visualization remain unstable: stochastic variational autoencoders inject sampling noise at inference, and methods reported on idiosyncratic cohorts defeat…
Authors not listed
The discovery of chemically novel or structurally anomalous metal-organic frameworks (MOFs) is essential for expanding reticular design space and enhancing dataset reliability. We present CHEM-AD (Chemically Unusual Metal–organic Frameworks via Autoencoder-based Detection), a label-free, CPU-efficient pipeline that…
Authors not listed
The automated discovery of chemical and catalytic reactions remains a major challenge in computational chemistry, particularly in complex systems where conventional methods struggle to identify optimal searching directions. Here, we propose Loxodynamics, a machine-learning-driven approach for reaction exploration via…
Authors not listed
The ability to generate crystal structures directly from textual descriptions marks a pivotal advancement in materials informatics and underscores the emerging role of large language models (LLMs) in inverse design. In this work, we introduce CrysText, a text-conditioned framework that generates crystal structures in…
Authors not listed
Computational chemistry has entered a new era where machine learning (ML) models—particularly graph neural networks and machine learning force fields—routinely deliver quantum mechanical accuracy at classical speeds, scaling to millions of atoms and reshaping workflows in drug discovery, catalysis, and materials…
Authors not listed
This work provides a rigorous theoretical investigation of selective error correction strategies for variational quantum algorithms, with focus on understanding the interplay between error suppression, circuit trainability, and computational resource requirements. We develop a mathematical framework that characterizes…
Authors not listed
Metastable states and the conformational transitions in between them are key to understanding dynamical behaviour and function of large-scale molecular systems. By combining basic dimensionality reduction techniques with a state-of-the art approximation of the Koopman operator associated to molecular dynamics…