Autoencoders are neural networks that compress data into a lower-dimensional space and reconstruct the original input, learning non-linear structures unlike linear PCA. Their two-stage design features an encoder that projects input data into a latent space to extract informative features.
This paper proposes a novel residual-centric approach for high-fidelity learned compression of scientific data. It introduces innovative residual coders designed to overcome limitations of existing methods, improving accuracy in high-tolerance regimes.
This content discusses the application of deep convolutional recurrent autoencoders to learn low-dimensional feature dynamics in fluid systems.
Industry 4.0 requires robust anomaly detection in Cyber-Physical Systems (CPS), but product-agnostic models have blind spots for subtle anomalies or targeted attacks. This work demonstrates this vulnerability and proposes a Product-Aware Autoencoder as a mitigation by restricting the learning domain.
This content describes research focused on improving reconstruction fidelity by utilizing a novel approach called JumpReLU SparseAutoencoders.