Results for "noise schedule"

Controls amount of noise added at each diffusion step.

Adjusting learning rate over training to improve convergence.

Diffusion model trained to remove noise step by step.

Variability introduced by minibatch sampling during SGD.

Generative model that learns to reverse a gradual noise process.

When a model fits noise/idiosyncrasies of training data and performs poorly on unseen data.

Expanding training data via transformations (flips, noise, paraphrases) to improve robustness.

Measures a model’s ability to fit random noise; used to bound generalization error.

Optimal estimator for linear dynamic systems.

Optimization under uncertainty.

Designing input features to expose useful structure (e.g., ratios, lags, aggregations), often crucial outside deep learning.

Number of samples per gradient update; impacts compute efficiency, generalization, and stability.

A gradient method using random minibatches for efficient training on large datasets.

A formal privacy framework ensuring outputs do not reveal much about any single individual’s data contribution.

Converting audio speech into text, often using encoder-decoder or transducer architectures.

Error due to sensitivity to fluctuations in the training dataset.

A narrow minimum often associated with poorer generalization.

Recovering training data from gradients.

Inferring sensitive features of training data.

Embedding signals to prove model ownership.

Learns the score (∇ log p(x)) for generative sampling.

Generator produces limited variety of outputs.

Two-network setup where generator fools a discriminator.

Monte Carlo method for state estimation.

Formal model linking causal mechanisms and variables.

Decomposes a matrix into orthogonal components; used in embeddings and compression.

Applying learned patterns incorrectly.

Software pipeline converting raw sensor data into structured representations.

Inferring the agent’s internal state from noisy sensor data.

Performance drop when moving from simulation to reality.