Discovering Influential Factors in Variational Autoencoders

A recent study has introduced a method for fully decentralized certified unlearning in machine learning, focusing on the removal of specific data influences from trained models without a central coordinator. This approach, termed RR-DU, employs a random-walk procedure to enhance privacy and mitigate data poisoning risks, providing convergence guarantees in convex scenarios and stationarity in nonconvex cases.

A novel Conditional Neural Cellular Automata (c-NCA) architecture has been proposed, enabling the generation of distinct topological structures, specifically MNIST digits, from a single seed. This approach emphasizes local interactions and translation equivariance, diverging from traditional generative models that rely on global reception fields.

A new reduced-space formulation for optimizing trained neural networks has been proposed, which evaluates the network's outputs and derivatives on a GPU. This method treats the neural network as a 'gray box,' leading to faster solves and fewer iterations compared to traditional full-space formulations. The approach has been demonstrated on two optimization problems, including adversarial generation for a classifier trained on MNIST images.

A recent study has introduced a pruned version of Capsule Networks (CapsNets), demonstrating that it can operate up to 9.90 times faster than traditional architectures by eliminating 95% of Primary Capsules while maintaining accuracy across various datasets, including MNIST and CIFAR-10.

Recent research has introduced geometry-aware noise injection techniques that enhance the training of machine learning models by considering the underlying structure of data. This approach involves projecting Gaussian noise onto the tangent space of a manifold and mapping it via geodesic curves, leading to improved model generalization and robustness.

A novel training objective called latent nonlinear denoising score matching (LNDSM) has been introduced, enhancing score-based generative models by integrating nonlinear dynamics with a VAE-based framework. This method reformulates the cross-entropy term using an approximate Gaussian transition, improving numerical stability and achieving superior sample quality on the MNIST dataset.