القائمة
组合性计算的理论框架 - 在抑制主导的阈值线性网络中使用低秩连接规则实现固定点的组合性，揭示大脑如何高效分解复杂任务为可复用的基本单元。
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
استنادا إلى تصنيف SOC المهني
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/hiyenwong/ai_collection --skill fixed-point-compositionality-via-low-rank-gluingيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
عرض SKILL.md
Graph-native Python reimplementation of the Information Dynamics of Music (IDyOM) model that represents predictive memories as explicit graph objects for musical expectation modeling and network analysis.
Physics-aware end-to-end deep reinforcement learning methodology for quadcopter control with actuator dynamics modeling.
Reinforced Dreamer methodology for asymmetric reinforcement learning using latent guidance to improve world model representations and behaviors in model-based RL.
| skill_id | fixed-point-compositionality-low-rank-gluing |
| name | Fixed Point Compositionality via Low-Rank Gluing |
| description | 组合性计算的理论框架 - 在抑制主导的阈值线性网络中使用低秩连接规则实现固定点的组合性，揭示大脑如何高效分解复杂任务为可复用的基本单元。 |
| version | 1.0.0 |
| author | Juliana Londono Alvarez |
| arxiv_id | 2606.07336v1 |
| categories | ["neuroscience","computational neuroscience","neural dynamics","theoretical neuroscience"] |
| tags | ["compositionality","low-rank networks","fixed points","threshold-linear networks","inhibition-dominated circuits","network modularity","neural assemblies"] |
| activation_keywords | ["compositionality","组合性","低秩连接","fixed point","固定点","threshold-linear","threshold linear","抑制主导","inhibition-dominated","neural modules","神经模块","task decomposition","任务分解"] |
| created_date | 2026-06-08T00:00:00.000Z |
| last_updated | 2026-06-08T00:00:00.000Z |
大脑在相对稳定结构和有限资源下生成高度灵活和复杂行为，其关键机制是组合性（compositionality）——将复杂任务分解为更简单、可复用的基本单元的能力。
核心思想：通过低秩权重矩阵连接不同的神经模块，使得固定点（fixed points）可以组合。
关键特性：
网络动力学：
dx/dt = -x + σ(Wx + b)
其中：
x：神经元活动向量W：低秩权重矩阵σ：阈值线性函数（ReLU-like）固定点条件：
x* = σ(Wx* + b)
数学表述：
给定两个固定点 x₁* 和 x₂*，其组合：
x_combined* = αx₁* + βx₂*
在特定条件下仍为稳定固定点，这依赖于：
import numpy as np
class CompositionalityNetwork:
"""
实现低秩组合性阈值线性网络
"""
def __init__(self, n_neurons, rank, threshold=0.0):
self.n = n_neurons
self.r = rank
self.threshold = threshold
# 低秩权重分解：W = U @ V.T
self.U = np.random.randn(n_neurons, rank) * 0.1
self.V = np.random.randn(n_neurons, rank) * 0.1
# 抑制主导：确保连接矩阵具有足够的抑制
self.W = self.U @ self.V.T
# 调整抑制权重使其主导
self.adjust_inhibition_dominance()
def adjust_inhibition_dominance(self, inhibition_ratio=0.7):
"""
调整权重矩阵使抑制连接主导
"""
# 随机选择抑制神经元
inhibition_mask = np.random.rand(self.n) < inhibition_ratio
self.V[inhibition_mask] *= -1.5 # 增强抑制
def threshold_linear(self, x):
"""阈值线性激活函数"""
return np.maximum(x - self.threshold, 0)
def compute_fixed_point(self, b, max_iter=1000, tol=1e-6):
"""
计算给定偏置下的固定点
"""
x = np.zeros(self.n)
for i in range(max_iter):
x_new = self.threshold_linear(self.W @ x + b)
if np.linalg.norm(x_new - x) < tol:
return x_new
x = x_new
return x
def compose_fixed_points(self, x1, x2, alpha=0.5, beta=0.5):
"""
组合两个固定点
"""
x_combined = alpha * x1 + beta * x2
# 检验是否为固定点
residual = np.linalg.norm(
x_combined - self.threshold_linear(self.W @ x_combined + np.zeros(self.n))
)
return x_combined, residual < 1e-4
def verify_compositionality(network, b1, b2):
"""
验证固定点的组合性
"""
# 计算两个独立固定点
fp1 = network.compute_fixed_point(b1)
fp2 = network.compute_fixed_point(b2)
# 组合
fp_combined, is_stable = network.compose_fixed_points(fp1, fp2)
print(f"Fixed Point 1: {np.linalg.norm(fp1):.4f}")
print(f"Fixed Point 2: {np.linalg.norm(fp2):.4f}")
print(f"Combined Fixed Point: {np.linalg.norm(fp_combined):.4f}")
print(f"Stability: {is_stable}")
return fp1, fp2, fp_combined, is_stable
组合性解释了大脑如何：
结构对应：
固定点作为记忆：
# 任务：序列决策分解为基本步骤
task_modules = {
'planning': compute_fixed_point(b_planning),
'execution': compute_fixed_point(b_execution),
'monitoring': compute_fixed_point(b_monitoring)
}
# 组合执行
combined_state = compose_fixed_points(
task_modules['planning'],
task_modules['execution']
)
低秩连接规则在抑制主导的阈值线性网络中实现了固定点的组合性，为理解大脑的组合性计算提供了理论基础。这一框架揭示了：