A neuroscience method to understanding the brain is to find and study the
preferred stimuli that highly activate an individual cell or groups of cells.
Recent advances in machine learning enable a family of methods to synthesize
preferred stimuli that cause a neuron in an artificial or biological brain to
fire strongly. Those methods are known as Activation Maximization (AM) or
Feature Visualization via Optimization. In this chapter, we (1) review existing
AM techniques in the literature; (2) discuss a probabilistic interpretation for
AM; and (3) review the applications of AM in debugging and explaining networks.

本章介绍了一种神经科学方法 ——Activation Maximization（AM），利用机器学习技术综合推理神经元的首选刺激，以大力激发人造或生物大脑中的单个细胞或细胞组，并讨论了现有 AM 技术的概率解释和其在网络调试和解释方面的应用。

透過特徵視覺化理解神經網絡：一項調查

Understanding Neural Networks via Feature Visualization: A survey

The densification and expansion of wireless network pose new challenges on
interference management and reducing energy consumption. This paper studies
energy-efficient resource management in heterogeneous networks by jointly
optimizing cell activation, user association and multicell multiuser channel
assignment, according to the long-term average traffic and channel conditions.
The proposed framework is built on characterizing the interference coupling by
pre-defined interference patterns, and performing resource allocation among
these patterns. In this way, the interference fluctuation caused by
(de)activating cells is explicitly taken into account when calculating the user
achievable rates. A tailored algorithm is developed to solve the formulated
problem in the dual domain by exploiting the problem structure, which gives a
significant complexity saving. Numerical results show a huge improvement in
energy saving achieved by the proposed scheme. The user association derived
from the proposed joint resource optimization is mapped to standard-compliant
cell selection biasing. This mapping reveals that the cell-specific biasing for
energy saving is quite different from that for load balancing investigated in
the literature.

本文研究了异构网络中的能效资源管理，通过联合优化单元激活、用户关联和多单元多用户信道分配来实现，同时考虑了长期平均流量和信道条件。研究提出了一种新型的干扰模式和资源分配算法，可大幅度提高能量节省效果，并显式考虑了由单元激活 / 去活导致的干扰波动，得到了标准兼容的单元选择偏差。其能源节省节省模式与负载平衡模式相比有明显的不同。