Uploaded January 2025 | Updated November 2026, 3 hours ago
Machine learning is being applied in computational physics to accelerate the solving of partial differential equations (PDEs). BuzzRobot guest Nick McGreivy from Princeton University found that 79% of studies claiming ML outperforms standard methods rely on weak baselines. He also uncovered widespread reporting biases, including outcome and publication biases, leading to an overly optimistic view of ML’s effectiveness in this field.
Timestamps:
0:00 Introduction
2:20 Machine learning for partial differential equations (PDE) research
3:30 Weak baselines
5:07 Baselines evaluate performance: what is a baseline, baselines in ML-for-PDE solving, concern over weak baselines
6:04 Methods: systematic review
7:53 Rule N1: make comparisons at equal accuracy or equal run time
9:17 Rule N2: compare to an efficient numerical method
10:07 Systematic review: results
13:03 Reporting biases
15:45 How positive is ML for PDE solving research?
19:03 Understanding causes
#machinelearning #machinelearningmodel #artificialintelligance #phyics #computationalphysics #deeplearning #ai #neuralnetworks #aiselfImprovement #aimodel #reinforcementlearning #llm #llms #airesearch #aichallenges #aiprogress #tech #techtalk #techtalks #aitalks #aitalk #science #python #pythonprogramming #ai #programming
Social Links:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-2s067rv7n-guPIMGe62rbp9ncxdnOUfQ
Machine learning is being applied in computational physics to accelerate the solving of partial differential equations (PDEs). BuzzRobot guest Nick McGreivy from Princeton University found that 79% of studies claiming ML outperforms standard methods rely on weak baselines. He also uncovered widespread reporting biases, including outcome and publication biases, leading to an overly optimistic view of ML’s effectiveness in this field.
Timestamps:
0:00 Introduction
2:20 Machine learning for partial differential equations (PDE) research
3:30 Weak baselines
5:07 Baselines evaluate performance: what is a baseline, baselines in ML-for-PDE solving, concern over weak baselines
6:04 Methods: systematic review
7:53 Rule N1: make comparisons at equal accuracy or equal run time
9:17 Rule N2: compare to an efficient numerical method
10:07 Systematic review: results
13:03 Reporting biases
15:45 How positive is ML for PDE solving research?
19:03 Understanding causes
#machinelearning #machinelearningmodel #artificialintelligance #phyics #computationalphysics #deeplearning #ai #neuralnetworks #aiselfImprovement #aimodel #reinforcementlearning #llm #llms #airesearch #aichallenges #aiprogress #tech #techtalk #techtalks #aitalks #aitalk #science #python #pythonprogramming #ai #programming
Social Links:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-2s067rv7n-guPIMGe62rbp9ncxdnOUfQ










