差别
这里会显示出您选择的修订版和当前版本之间的差别。
| 两侧同时换到之前的修订记录 前一修订版 后一修订版 | 前一修订版 | ||
| 深度学习:深度学习概述 [2026/03/02 20:58] – [计算题] 张叶安 | 深度学习:深度学习概述 [2026/03/02 21:00] (当前版本) – [计算题答案] 张叶安 | ||
|---|---|---|---|
| 行 387: | 行 387: | ||
| 11. 一个神经元有四个输入$x_1=1.0$、$x_2=-0.5$、$x_3=0.3$、$x_4=-0.8$,权重$w_1=0.5$、$w_2=0.3$、$w_3=-0.2$、$w_4=0.4$,偏置$b=-0.1$。请计算: | 11. 一个神经元有四个输入$x_1=1.0$、$x_2=-0.5$、$x_3=0.3$、$x_4=-0.8$,权重$w_1=0.5$、$w_2=0.3$、$w_3=-0.2$、$w_4=0.4$,偏置$b=-0.1$。请计算: | ||
| + | |||
| (1) 加权和$z$ | (1) 加权和$z$ | ||
| 行 430: | 行 431: | ||
| 11. **解答**: | 11. **解答**: | ||
| | | ||
| - | | + | (1) 加权和: |
| - | $$z = 0.5 \times 1.0 + 0.3 \times (-0.5) + (-0.2) \times 0.3 + 0.4 \times (-0.8) - 0.1$$ | + | |
| - | $$z = 0.5 - 0.15 - 0.06 - 0.32 - 0.1 = -0.13$$ | + | $$z = 0.5 \times 1.0 + 0.3 \times (-0.5) + (-0.2) \times 0.3 + 0.4 \times (-0.8) - 0.1$$ |
| + | |||
| + | $$z = 0.5 - 0.15 - 0.06 - 0.32 - 0.1 = -0.13$$ | ||
| | | ||
| - | | + | (2) Tanh输出: |
| - | $$y = \tanh(-0.13) = \frac{e^{-0.13} - e^{0.13}}{e^{-0.13} + e^{0.13}} \approx \frac{0.878 - 1.139}{0.878 + 1.139} \approx -0.129$$ | + | |
| + | $$y = \tanh(-0.13) = \frac{e^{-0.13} - e^{0.13}}{e^{-0.13} + e^{0.13}} \approx \frac{0.878 - 1.139}{0.878 + 1.139} \approx -0.129$$ | ||
| | | ||
| - | | + | (3) Softmax概率: |
| - | 三个输入值:0.5、-0.13(当前神经元)、0.3 | + | |
| - | $$\text{分母} = e^{0.5} + e^{-0.13} + e^{0.3} = 1.649 + 0.878 + 1.350 = 3.877$$ | + | 三个输入值:0.5、-0.13(当前神经元)、0.3 |
| - | $$P(\text{第2类}) = \frac{e^{-0.13}}{3.877} = \frac{0.878}{3.877} \approx 0.226$$ | + | |
| + | $$\text{分母} = e^{0.5} + e^{-0.13} + e^{0.3} = 1.649 + 0.878 + 1.350 = 3.877$$ | ||
| + | |||
| + | $$P(\text{第2类}) = \frac{e^{-0.13}}{3.877} = \frac{0.878}{3.877} \approx 0.226$$ | ||
| 12. **解答**: | 12. **解答**: | ||
| | | ||
| - | | + | **样本1**:(1, |
| - | - 计算:$0.1 \times 1 + (-0.1) \times 2 + 0 = 0.1 - 0.2 = -0.1 < 0$,预测负类(错误) | + | - 计算:$0.1 \times 1 + (-0.1) \times 2 + 0 = 0.1 - 0.2 = -0.1 < 0$,预测负类(错误) |
| - | - 更新:$w_1 = 0.1 + 0.5 \times 1 = 0.6$,$w_2 = -0.1 + 0.5 \times 2 = 0.9$,$b = 0 + 0.5 = 0.5$ | + | - 更新:$w_1 = 0.1 + 0.5 \times 1 = 0.6$,$w_2 = -0.1 + 0.5 \times 2 = 0.9$,$b = 0 + 0.5 = 0.5$ |
| | | ||
| - | | + | **样本2**:(-1, |
| - | - 计算:$0.6 \times (-1) + 0.9 \times (-1) + 0.5 = -0.6 - 0.9 + 0.5 = -1.0 < 0$,预测负类(正确) | + | - 计算:$0.6 \times (-1) + 0.9 \times (-1) + 0.5 = -0.6 - 0.9 + 0.5 = -1.0 < 0$,预测负类(正确) |
| - | - 不更新 | + | - 不更新 |
| | | ||
| - | | + | **样本3**:(2, |
| - | - 计算:$0.6 \times 2 + 0.9 \times 1 + 0.5 = 1.2 + 0.9 + 0.5 = 2.6 > 0$,预测正类(正确) | + | - 计算:$0.6 \times 2 + 0.9 \times 1 + 0.5 = 1.2 + 0.9 + 0.5 = 2.6 > 0$,预测正类(正确) |
| - | - 不更新 | + | - 不更新 |
| | | ||
| - | | + | **一轮迭代后参数**:$w_1 = 0.6$,$w_2 = 0.9$,$b = 0.5$ |
| - | --- | ||
| - | **本章完** | ||