arXiv subjects

Tyler Clark

Publications and source records attributed to Tyler Clark.

5 recordsLinked to original sources

arXiv2025

Recurrent Off-Policy Deep Reinforcement Learning Doesn't Have to be Slow

Recurrent off-policy deep reinforcement learning models achieve state-of-the-art performance but are often sidelined due to their high computational demands. In response, we introduce RISE (Recurrent Integration via Simplified Encodings), a novel approach that can leverage recurrent networks in any image-based off-policy RL setting without significant computational overheads via using both learnable and non-learnable encoder layers. When integrating RISE into leading non-recurrent off-policy RL algorithms, we observe a 35.6% human-normalized interquartile mean (IQM) performance improvement across the Atari benchmark. We analyze various implementation strategies to highlight the versatility and potential of our proposed framework.

cs.LG↗

arXiv2024

Beyond The Rainbow: High Performance Deep Reinforcement Learning on a Desktop PC

Rainbow Deep Q-Network (DQN) demonstrated combining multiple independent enhancements could significantly boost a reinforcement learning (RL) agent's performance. In this paper, we present "Beyond The Rainbow" (BTR), a novel algorithm that integrates six improvements from across the RL literature to Rainbow DQN, establishing a new state-of-the-art for RL using a desktop PC, with a human-normalized interquartile mean (IQM) of 7.4 on Atari-60. Beyond Atari, we demonstrate BTR's capability to handle complex 3D games, successfully training agents to play Super Mario Galaxy, Mario Kart, and Mortal Kombat with minimal algorithmic changes. Designing BTR with computational efficiency in mind, agents can be trained using a high-end desktop PC on 200 million Atari frames within 12 hours. Additionally, we conduct detailed ablation studies of each component, analyzing the performance and impact using numerous measures. Code is available at https://github.com/VIPTankz/BTR.

cs.AI↗

arXiv2024

The N-Gene Model for Evolutionary Games

In this study, we develop a novel evolutionary model that incorporates Mendelian genetics, continuous strategies, and the potential for multiple genes to contribute to a single phenotypic trait. The evolution of altruistic behavior, which confers benefits to others at a cost to the individual, remains a fundamental question in evolutionary biology. While previous models have investigated the conditions favoring the emergence and stability of altruism, they have often employed simplifying assumptions, such as single-gene inheritance and discrete strategies. We employ a modified dictator game as the framework for evolutionary interactions and explore the stability of altruistic behavior under various conditions. Our primary result demonstrates that when considering heterozygous genes, altruism can be evolutionarily stable at cost-to-benefit ratios exceeding unity, even with initially low frequencies of altruists in the population. This finding contrasts with the case of homozygous genes, where altruism is only stable at cost-to-benefit ratios greater than 2. The generality of our approach allows for its application to a diverse range of evolutionary games and interactions, providing a powerful tool for investigating the emergence and maintenance of social behaviors and personality traits. Our results contribute to the understanding of the evolutionary mechanisms underlying altruism and underscore the importance of incorporating genetic complexity in evolutionary models. This work has implications for the study of social evolution and the genetic architecture of complex behavioral phenotypes.

q-bio.PE↗

arXiv2017

Quasinormal modes and absorption cross sections of Born-Infeld-de Sitter black holes

In this paper, we have studied QNM modes and absorption cross sections of Born-Infeld-de Sitter black holes. WKB approximation is employed to compute the QNM modes of massless scalar fields. We have also used null geodesics to compute quasi-normal modes in the eikonal approximation. In the eikonal limit QNMs of black holes are determined by the parameters of the circular null geodesics. Unstable circular null orbits are derived from the effective metric which is obeyed by light rays under the influence of a nonlinear electromagnetic field. Comparison is shown with the QNM of the linear electromagnetic counterpart, the Reissner-Nordstr\"{o}m black hole. Furthermore, the null geodesics are employed to compute the absorption cross sections in the high frequency limit via the sinc approximation.

gr-qc↗

arXiv2014

Black holes in massive gravity: Quasi-normal modes of scalar perturbations

We have studied quasinormal modes of scalar perturbations of a black hole in massive gravity. The parameters of the theory, such as the mass of the black hole, the scalar charge of the black hole and the spherical harmonic index is varied to see how the corresponding quasinormal frequencies change. We have also studied the massive scalar field perturbations. Most of the work is done using WKB approach while sections are devoted to compute quasinormal modes via the unstable null geodesics approach and the Poschl-Teller approximation. Comparisons are done with the Schwarzschild black hole.

gr-qc↗