本頁只刊出中文翻譯與中文說明；英文原文請見下方原文連結。
原文連結
論文資訊
- 類型：已發表論文
- 日期：2025-04-11
摘要
Physical 網絡s can develop tuned responses, or functions, by design, by 演化, or by learning via local rules. In all of these cases, tunable degrees of freedom characterizing internal interactions are modified to lower a cost penalizing deviations from desired outputs. An important class of such 網絡s follows dynamics that minimize a global physical quantity, or 李雅普諾夫 function, with respect to physical degrees of freedom. In such 網絡s, learning is a "double optimization" process in which two quantities, one defined by the task and the other prescribed by physics, are minimized with respect to different but coupled sets of variables. Here, we show how this learning process couples the high-dimensional "cost landscape" to the "physical landscape," linking the physical and cost Hessian matrices. Phy
※ 此為已發表論文，全文需透過期刊付費取得