Results 101 to 110 of about 119,888 (330)

High‐Fidelity Synthetic Data Replicates Clinical Prediction Performance in a Million‐Patient Diabetes Cohort

Advanced Science, EarlyView.
This study generates high‐fidelity synthetic longitudinal records for a million‐patient diabetes cohort, successfully replicating clinical predictive performance. However, deeper analysis reveals algorithmic biases and trajectory inconsistencies that escape standard quality metrics. These findings challenge current validation norms, demonstrating why a
Francisco Ortuño, Víctor M. de la Oliva Roque, Javier‐Ignacio Ramirez‐Lopez, David P. Kreil, Joaquín Dopazo, Carlos Loucera   +5 more
wiley   +1 more source

Zero-bias autoencoders and the benefits of co-adapting features [PDF]

, 2015
Regularized training of an autoencoder typically results in hidden unit biases that take on large negative values. We show that negative biases are a natural result of using a hidden layer whose responsibility is to both represent the input data and act ...
Konda, Kishore, Krueger, David, Memisevic, Roland   +2 more
core  

Processsing Simple Geometric Attributes with Autoencoders

, 2019
Image synthesis is a core problem in modern deep learning, and many recent architectures such as autoencoders and Generative Adversarial networks produce spectacular results on highly complex data, such as images of faces or landscapes.
Almansa, Andrés, Gousseau, Yann, Ladjal, Saïd, Newson, Alasdair   +3 more
core   +3 more sources

Solid Harmonic Wavelet Bispectrum for Image Analysis

Advanced Science, EarlyView.
The Solid Harmonic Wavelet Bispectrum (SHWB), a rotation‐ and translation‐invariant descriptor that captures higher‐order (phase) correlations in signals, is introduced. Combining wavelet scattering, bispectral analysis, and group theory, SHWB achieves interpretable, data‐efficient representations and demonstrates competitive performance across texture,
Alex Brown, Mathilda Avirett‐Mackenzie, Carolin Villforth, Georgios Exarchakis   +3 more
wiley   +1 more source

Deep Learning Autoencoders for Fast Fourier Transform-Based Clustering and Temporal Damage Evolution in Acoustic Emission Data from Composite Materials

Infrastructures
Structural health monitoring (SHM) in fiber-reinforced polymer (FRP) composites is essential to ensure safety and reliability during service, particularly in critical industries such as aerospace and wind energy. Traditional methods of analyzing Acoustic
Serafeim Moustakidis, Konstantinos Stergiou, Matthew Gee, Sanaz Roshanmanesh, Farzad Hayati, Patrik Karlsson, Mayorkinos Papaelias   +6 more
doaj   +1 more source

Gaussian Process Prior Variational Autoencoders [PDF]

, 2018
Variational autoencoders (VAE) are a powerful and widely-used class of models to learn complex data distributions in an unsupervised fashion. One important limitation of VAEs is the prior assumption that latent sample representations are independent and ...
Casale, Francesco Paolo, Dalca, Adrian V, Fusi, Nicolo, Listgarten, Jennifer, Saglietti, Luca   +4 more
core   +1 more source

Toroidal AutoEncoder

CoRR, 2019
Enforcing distributions of latent variables in neural networks is an active subject. It is vital in all kinds of generative models, where we want to be able to interpolate between points in the latent space, or sample from it. Modern generative AutoEncoders (AE) like WAE, SWAE, CWAE add a regularizer to the standard (deterministic) AE, which allows to ...
Maciej Mikulski, Jaroslaw Duda 0001
openaire   +2 more sources

Decoding Naturalistic Episodic Memory with Artificial Intelligence and Brain‐Machine Interface

Advanced Science, EarlyView.
Episodic memory weaves together what, where, and when of experience into a personal narrative. Cutting‐edge AI models may decode this intricate process in real‐life settings, revealing how neural activity encodes naturalistic memories. By merging AI with brain–machine interfaces, researchers are edging closer to mapping and even engineering memory ...
Dong Song
wiley   +1 more source

Improving the explainability of autoencoder factors for commodities through forecast-based Shapley values

Scientific Reports
Autoencoders are dimension reduction models in the field of machine learning which can be thought of as a neural network counterpart of principal components analysis (PCA).
Roy Cerqueti, Antonio Iovanella, Raffaele Mattera, Saverio Storani   +3 more
doaj   +1 more source

Autoencoding any Data through Kernel Autoencoders

CoRR, 2018
This paper investigates a novel algorithmic approach to data representation based on kernel methods. Assuming that the observations lie in a Hilbert space X, the introduced Kernel Autoencoder (KAE) is the composition of mappings from vector-valued Reproducing Kernel Hilbert Spaces (vv-RKHSs) that minimizes the expected reconstruction error.
Pierre Laforgue, Stéphan Clémençon, Florence d'Alché-Buc   +2 more
openaire   +3 more sources