Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Reinforcement Patching (ReinPatch) is the first framework to jointly optimize a sequence patching policy and its downstream sequence backbone model using reinforcement learning.method 95%
• ReinPatch bypasses the need for continuous relaxations and performs dynamic patching policy optimization by formulating patch boundary placement as a discrete decision process optimized via Group Relative Policy Gradient (GRPG).method 95%
• ReinPatch allows strict enforcement of a desired compression rate.technical 90%
• ReinPatch demonstrates compelling performance compared to state-of-the-art data-driven patching strategies on time-series forecasting datasets.result 90%
• The ReinPatch patching module can be extracted as a standalone foundation patcher.technical 90%
• On the ETTh1 dataset, ReinPatch achieves an average performance of 0.424±0.008 for a prediction length of 192.result 85%
• On the ETTm2 dataset, ReinPatch achieves an average performance of 0.321±0.003 for a prediction length of 720.result 85%
• Reinforcement Patching (ReinPatch) is the first framework to jointly optimize a sequence patching policy and its downstream sequence backbone model using reinforcement learning.method 95%
• ReinPatch bypasses the need for continuous relaxations and performs dynamic patching policy optimization by formulating patch boundary placement as a discrete decision process optimized via Group Relative Policy Gradient (GRPG).method 95%
• ReinPatch allows strict enforcement of a desired compression rate.technical 90%
• ReinPatch demonstrates compelling performance compared to state-of-the-art data-driven patching strategies on time-series forecasting datasets.result 90%
• The ReinPatch patching module can be extracted as a standalone foundation patcher.technical 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card