Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 5 of 5 for “"Gradient clipping"”.

  1. Understanding Adaptivity in Machine Learning Optimization: Theories and Algorithms

    … attacks. In the third part, we study the gradient clipping operation for differentially private SGD. Gradient clipping adds a form of adaptation to SGD that can potentially hurt convergence. We identify regimes where gradient clipping is not an issue and verify the existence of these …

    umn Repository record for Understanding Adaptivity in Machine Learning Optimization: Theories and Algorithms (opens in a new tab)

  2. Generalization in Machine Learning Through Information-Theoretic Lens

    … analysis for models trained using stochastic gradient descent (SGD). We do so by invoking an auxiliary weight process and by approximating SGD using stochastic differential equations (SDE). Our analysis reveals intriguing phenomena such as epoch-wise double descent of gradient dispersion when …

    ottawa-retro Repository record for Generalization in Machine Learning Through Information-Theoretic Lens (opens in a new tab)

  3. Unconstrained online convex optimization: regularization in the adversarial setting and beyond

    … to an optimal solution, and receive gradient feedback that is stochastic, corrupted, or both. This thesis studies whether these two difficulties can be resolved simultaneously: we develop online convex optimization algorithms on an unconstrained domain that adapt to the unknown …

    bu Repository record for Unconstrained online convex optimization: regularization in the adversarial setting and beyond (opens in a new tab)

  4. Adaptive strategies in non-convex optimization

    … optimizer used for such scenarios is Stochastic Gradient Descent (SGD), but its performance depends crucially on the choice of its step sizes. Tuning of step sizes is notoriously laborious and the optimal choice can vary drastically across different problems. To save the labor of tuning, adaptive …

    bu Repository record for Adaptive strategies in non-convex optimization (opens in a new tab)

  5. PhD

    … to privacy attacks. Membership inference and gradient-based reconstruction attacks have demonstrated that, without explicit constraints, deep learning models can encode private features within their parameters. Consequently, Differential Privacy (DP) has emerged as the formal mathematical …

    anu Repository record for PhD (opens in a new tab)