The (1 + β)-Process

The \((1+\beta )\)-process is a process mixing the \(\TwoChoice \) process (with probability \(\beta \in (0, 1]\)) and the \(\OneChoice \) process (with probability \(1 - \beta \)). More formally, it can be defined as follows.

The \((1+\beta )\)-process:
Iteration: At each step \(t = 1, 2, \ldots \),

  • • Sample two bins \(i_1 = i_1^t\) and \(i_2 = i_2^t\) uniformly at random.

  • • With probability \(\beta \) allocate to bin \(j \in \{ i_1, i_2 \}\) such that \(x_{j}^t = \min \{ x_{i_1}^t, x_{i_2}^t \}\).

  • • Otherwise, allocate to bin \(i_1\).

The process was introduced by Mitzenmacher (1999) as a model of \(\TwoChoice \) with erroneous comparisons. Peres, Talwar and Wieder (2015) proved that this process has w.h.p. an \(\Oh (\frac {\log n}{\beta })\) gap for \(\beta \in (0, 1]\). This process is also known to work well in the \(\Batched \) setting.

Your browser does not support the HTML5 canvas tag.

.

References

  • [1] M. Mitzenmacher. “On the analysis of randomized load balancing schemes”. In: Theory Comput. Syst. 32.3 (1999), pp. 361–386. issn: 1432-4350. doi: 10.1007/s002240000122.

  • [2] Yuval Peres, Kunal Talwar, and Udi Wieder. “Graphical balanced allocations and the (1 + β)-choice process”. In: Random Structures & Algorithms 47.4 (2015), pp. 760–775. issn: 1042-9832. doi: 10.1002/rsa.20558.