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 9 of 9 for “"Zeroth order optimization"”.
-
Design of a Current Mode Circuit for Zeroth-Order Optimization of Memristor-Based Neuromorphic Systems
… perturbation algorithm (WSSPA). WSSPA is a zeroth-order ANN training algorithm that estimates loss gradients using perturbations of neuron inputs, avoiding large-overhead circuitry needed for backpropagation. The training circuit optimizes the ANN by adjusting the conductance of the …
-
TOWARDS IMPROVING GENERALIZATION OF DEEP NEURAL NETWORKS: DATA, PARAMETER OPTIMIZATION, AND DEPLOYMENT
… including training data preparation, parameter optimization, and model deployment. From the data perspective, I propose Concurrent Adversarial Learning (ConAdv), which reduces the computational cost of adversarial training in large-batch settings. By smoothing the decision boundary more …
-
Efficient Deep Neural Network Pruning: From Convolutional Networks to Large Language Models
… method called Influence Function based Second-Order (IFSO). Our findings show significant differences in loss changes with and without retraining, with the latter resulting in worse Top-1 accuracy. To address this issue computationally efficiently, we propose a closed-form estimator for true …
-
Enabling Conditional Generation Without Training: Plug & Play Generative Modelling using Diffusion Models
… which replaces the need for separate models with optimization iterations embedded within the generation process. Third, we enhance Steered Diffusion with DreamGuider, a method that removes its memory overhead and reliance on handcrafted parameters. By incorporating zeroth-order optimization and …
-
Optimizing Microservices Autoscaling
… efficiency. The reports delves into a variety of zeroth-order optimization techniques integrated with Bayesian optimization and compressive sensing. This approach aims to outperform traditional stochastic gradient descent methods in efficiency and scalability. Our investigations of these …
-
Adversarial Machine Learning and Defenses for Automated and Connected Vehicles
… black-box adversarial attacks executed through zeroth-order optimization methods like ZO-SignSGD in a lane-changing scenario. The research first train the policies with finely turned hyper-parameters in the lane-changing environment and achieving a high performance. With a good policy as a base, …
-
Stochastic modelling and simulation based optimization of specialty fibres for high power fibre lasers
… and applied mathematics into a simulation-based optimization environment. The proposed computational and mathematical tools are synergistically combined to produce globally optimized specialty fibres for high power fibre lasers. The main contributions to new knowledge made by this work are …
-
A NEW ZEROTH-ORDER ORACLE FOR DISTRIBUTED AND NON-STATIONARY LEARNING
<p>Zeroth-Order (ZO) methods have been applied to solve black-box or simulation-based optimization prroblems. These problems arise in many important applications nowa- days, e.g., generating adversarial attacks on machine learning systems, learning to control the system with complicated physics …
-
Ανάπτυξη και εφαρμογές του αλγόριθμου S.Q.A. για την επίλυση προβλημάτων παραμετρικής βελτιστοποίησης της απόκρισης εύκαμπτων μη - γραμμικών δυναμικών συστημάτων
… particular criteria, as well as for the overall optimization problem formulation. Consequently, many researchers in the field resort to efficient non-conventional solution techniques, including the implementation of various parameter optimization methodologies. The current response optimization …