Robust consideration of importance sampling in digital communication system

Importance sampling techniques for estimating the bit error rate in digital communication systems

“…First, we can see that the optimal setting happens when % 1=2 2 , which is consistent with the results in [1]. Second, the results between Rayleigh tail and Rayleigh þ uniform disproves the conjecture from [13] that a mixed biasing p.d.f. of any tail-like biasing distribution with a uniform distribution should always be more robust than the initial tail-like biasing distribution.…”

“…involving a Rayleigh tail and the uniform, and an exponential tail and the uniform. Our results disprove the conjecture from [13] that a mixed biasing p.d.f. of any tail-like biasing distribution with a uniform distribution should always be more robust than the initial tail-like biasing distribution.…”

“…when the threshold follows a uniform distribution. Table 1 corrects some typographical errors in [13]. Table 2 shows that for the three choices of ðc 1 , c 2 Þ, the IS estimator with the mixed Rayleigh biasing p.d.f.…”

“…Shih and Song [13] combine a Gaussian tail and a uniform as a ''mixed'' biasing p.d.f., and they show that the mixed certain biasing p.d.f. performs better than the Gaussian tail distribution.…”

Importance Sampling Techniques for Estimating The Bit Error Rate In Digital Communication Systems

“…Shih and Song (1995) combines a Gaussian tail and a uniform distribution as a "mixed" biasing p.d.f., and they show that the mixed performs better than does the Gaussian tail distribution alone. This paper reviews, expands upon, and evaluates the three well-known biasing p.d.f.s for estimating the BER, including some simple "fixes" to the mixed biasing p.d.f., e.g., the Gaussian tail and the uniform.…”