Theoretical Computer Science ( IF 1.1 ) Pub Date : 2021-04-30 , DOI: 10.1016/j.tcs.2021.04.026 Di Yan , Yu Yu , Hanlin Liu , Shuoyao Zhao , Jiang Zhang
We revisit the Learning Sparse Parities with Noise (LSPN) problem on k out of n variables for , and present the following findings.
- 1.
For true parity size for any , and noise rate , the first algorithm solves the (n,k,η)-LSPN problem with constant probability and time/sample complexity .
- 2.
For any , , and , our second algorithm solves the (n,k,η)-LSPN problem with constant probability and time/sample complexity .
- 3.
We show a “win-win” result about reducing the number of samples. If there is an algorithm that solves -LSPN problem with probability , time/sample complexity for , any noise rate and . Then, either there exists an algorithm that solves the -LSPN problem under lower noise rate using only 2n samples, or there exists an algorithm that solves the -LSPN problem for a much larger with probability , and time complexity , using only n samples.
中文翻译:
噪声存在下稀疏奇偶校验的一种改进算法
我们重温学习稀疏平价与噪声(LSPN)问题ķ出ñ变量,并提出以下发现。
- 1。
真正的奇偶校验大小 对于任何 和噪音率 ,第一种算法以恒定的概率和时间/样本复杂度解决了(n,k,η)-LSPN问题。
- 2。
对于任何 , , 和 ,我们的第二种算法以恒定的概率和时间/样本复杂度解决了(n,k,η)-LSPN问题。
- 3。
我们展示了减少样本数量的“双赢”结果。如果有一种算法可以解决-LSPN问题的概率 ,时间/样本复杂度 为了 ,任何噪音率 和 。然后,要么存在一种可以解决以下问题的算法-低噪声率下的LSPN问题 仅使用2 n个样本,或者存在一种算法可以解决-LSPN问题更大 很有可能 和时间复杂度 ,仅使用n个样本。