Exploring the Performance of Pruning Methods in Neural Networks: An Empirical Study of the Lottery Ticket Hypothesis
Authors:
Eirik Fladmark,
Muhammad Hamza Sajjad,
Laura Brinkholm Justesen
Abstract:
In this paper, we explore the performance of different pruning methods in the context of the lottery ticket hypothesis. We compare the performance of L1 unstructured pruning, Fisher pruning, and random pruning on different network architectures and pruning scenarios. The experiments include an evaluation of one-shot and iterative pruning, an examination of weight movement in the network during pru…
▽ More
In this paper, we explore the performance of different pruning methods in the context of the lottery ticket hypothesis. We compare the performance of L1 unstructured pruning, Fisher pruning, and random pruning on different network architectures and pruning scenarios. The experiments include an evaluation of one-shot and iterative pruning, an examination of weight movement in the network during pruning, a comparison of the pruning methods on networks of varying widths, and an analysis of the performance of the methods when the network becomes very sparse. Additionally, we propose and evaluate a new method for efficient computation of Fisher pruning, known as batched Fisher pruning.
△ Less
Submitted 26 March, 2023;
originally announced March 2023.
Looking for Stable Celestial Systems Using Bayesian Optimisation
Authors:
Eirik Fladmark,
Teodora Reu,
Laura Brinkholm Justesen
Abstract:
This paper presents a study of the use of numerical simulation and Bayesian optimisation techniques to investigate the dynamics of celestial systems. Initially, the study focuses on Lagrange points in restricted three-body systems where a 2D three-body system simulator is employed to locate the five Lagrange points. An appropriate loss function is developed to capture the gravitational stability o…
▽ More
This paper presents a study of the use of numerical simulation and Bayesian optimisation techniques to investigate the dynamics of celestial systems. Initially, the study focuses on Lagrange points in restricted three-body systems where a 2D three-body system simulator is employed to locate the five Lagrange points. An appropriate loss function is developed to capture the gravitational stability of the system, and the stability properties of the different Lagrange points are explored. Additionally, the study investigates how varying the number of variables for the satellite impacts the search for the Lagrange points. Finally, the scope of the study is expanded to explore stable configurations in multi-star systems represented by regular convex n-gons. In this case, Bayesian optimisation is used to find suitable settings for the n-gon's radius and the stars' velocity vectors, such that the overall system is stable.
△ Less
Submitted 26 March, 2023;
originally announced March 2023.
