b****y 发帖数: 257 | 1 Dear Da niu,
I am trying to compare the performance of several algorithms that has
different succ rates and use
different number of resources. For example, algo A has overall succ rate of
60% while using 10 resources,
algo B has overal succ rate of 85% while using 30 resources. It's hard to
compare them since algo A has
less succ rate but use less resources. I am thinking of normalize their
success rate to the same(i.e. 95%) by
assuming each run will have the same succ probability, i.e each run | b****y 发帖数: 257 | 2 please let me know if I didn't describe my question clearly!!
of
Algo
【在 b****y 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Dear Da niu, : I am trying to compare the performance of several algorithms that has : different succ rates and use : different number of resources. For example, algo A has overall succ rate of : 60% while using 10 resources, : algo B has overal succ rate of 85% while using 30 resources. It's hard to : compare them since algo A has : less succ rate but use less resources. I am thinking of normalize their : success rate to the same(i.e. 95%) by : assuming each run will have the same succ probability, i.e each run
| b****y 发帖数: 257 | 3 please tell me if it is not possible too!!
of
Algo
【在 b****y 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Dear Da niu, : I am trying to compare the performance of several algorithms that has : different succ rates and use : different number of resources. For example, algo A has overall succ rate of : 60% while using 10 resources, : algo B has overal succ rate of 85% while using 30 resources. It's hard to : compare them since algo A has : less succ rate but use less resources. I am thinking of normalize their : success rate to the same(i.e. 95%) by : assuming each run will have the same succ probability, i.e each run
| b****y 发帖数: 257 | 4 bump
of
Algo
【在 b****y 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Dear Da niu, : I am trying to compare the performance of several algorithms that has : different succ rates and use : different number of resources. For example, algo A has overall succ rate of : 60% while using 10 resources, : algo B has overal succ rate of 85% while using 30 resources. It's hard to : compare them since algo A has : less succ rate but use less resources. I am thinking of normalize their : success rate to the same(i.e. 95%) by : assuming each run will have the same succ probability, i.e each run
| s*****n 发帖数: 2174 | 5 你这问题问的本身就很模糊, 根本看不明白.
什么是 resource, 什么是 N, 你这个实验
到底是run in parrellel 还是 sequential.
你一开始所谓的 succ rate 应该指的是单次的成功率.
可是最后考虑的succ rate, 又是sequential experiments
恰好在第N+1次首次成功的概率. 这都是完全不相关
的概念.
如果就是想 知道 什么样的N能够使得
(1-a)^N * a = 0.95,
那么取一下log不就行了吗? |
|