Basics of Neural Network Programming - More derivatives examples

最新推荐文章于 2021-06-22 18:57:22 发布

王彩旗 edwardwangcq.com

最新推荐文章于 2021-06-22 18:57:22 发布

阅读量60

点赞数

分类专栏：人工智能 # Neural Networks and Deep Learning

本文链接：https://blog.csdn.net/edward_wang1/article/details/117989204

版权

142 篇文章 0 订阅

订阅专栏

32 篇文章 0 订阅

订阅专栏

Let's see more complex examples where the slope of the function can be different at different points of the function.

Figure-1 shows the plot of function $f(a)=a^{2}$ . Let's look at some points:

If $a=2$ , then $f(a)=4$
Nudge $a$ slightly to the right, thus $a=2.001$ , then $f(a)\approx 4.004$
The slope at point $a=2$ is $\frac{\Delta y}{\Delta a}=\frac{0.004}{0.001}=4$
Formally $\frac{df(a)}{da}|_{a=2}=4$
If $a=5$ , then $f(a)=25$
Nudge $a$ slightly to the right, thus $a=5.001$ , then $f(a)\approx 25.01$
The slope at point $a=5$ is $\frac{\Delta y}{\Delta a}=\frac{0.01}{0.001}=10$
Formally $\frac{df(a)}{da}|_{a=5}=10$

So, the mathematical definition is like this:

$\frac{df(a)}{da}=\frac{d}{da}f(a)=\frac{d}{da}a^{2}=2a$

Another example:

$f(a)=a^{3}$

Then $\frac{d}{da}f(a)=3a^{2}$

If $a=2$ , then $f(a)=8$
Nudge $a$ slightly to $a=2.001$ ,
Because $\frac{d}{da}f(a)=12|_{a=2}$ , it means $\Delta f(a)=12\times\Delta a=12\times 0.001=0.012$
then $f(a)=8.012$

Last example:

$f(a)=\ln (a)$

Then $\frac{d}{da}f(a)=\frac{d}{da}ln(a)=\frac{1}{a}$

If $a=2$ , then $f(a)=ln(a)=ln(2)\approx 0.69315$
When $a=2.001$ , because $\frac{d}{da}f(a)=\frac{1}{2}|_{a=2}$ , it means $\Delta f(a)=0.5\times\Delta a=0.5\times 0.001=0.0005$
then $f(a)=0.69315+0.0005=0.69365$

So, two take aways for this class:

The derivative of a function is just the slope of the function. And the slope of the function can be different at different points
If you look up the derivative of a function, you can flip open your calculus textbook or look at Wikipedia, you can often get a formula for the slope of these functions at different points

<end>

关注

专栏目录