神经网络架构示意图

一幅多层前馈神经网络扁平技术示意图，从左到右排列。五层垂直结构：输入层（4 个珊瑚色圆形节点）、两层隐藏层（各 6 个靛青色节点）、输出层（3 个薄荷色节点）。相邻层间所有节点以浅灰色细边连接；琥珀色高亮边展示一条从输入节点 2 到输出节点 1 的前向传播路径。节点直径按激活值大小比例显示，图例说明此视觉提示。图下方：40×40 pt 缩略图框中展示 ReLU 和 Sigmoid 激活函数曲线，各自标注名称。白色背景，纯色，8 pt 等宽标签，1 pt 规则线，适合会议幻灯片风格。

A clean flat technical diagram of a multi-layer feedforward neural network, arranged left to right. Five vertical layers: input layer (4 nodes, coral circles) labeled "Input Layer" at top; two hidden layers (6 nodes each, indigo circles) labeled "Hidden Layer 1" and "Hidden Layer 2"; output layer (3 nodes, mint circles) labeled "Output Layer". Every node in adjacent layers is connected by thin light-grey edges. Selected edges are highlighted in amber to illustrate a single forward-pass path from input node 2 to output node 1. Node diameters are proportional to activation magnitude — a visual cue shown in the legend. Below the diagram, activation function mini-plots: small ReLU and Sigmoid curves in 40×40 pt thumbnail boxes labeled "ReLU" and "Sigmoid". White background, flat colors, 8 pt monospace labels, 1 pt rule lines. Layout designed for a conference slide.