The σ-Noisy-Load Process

The \(\SigmaNoisyLoad \) process is a noisy version of \(\TwoChoice \) where the load estimates are perurbed by a Gaussian distribution \(\mathcal {N}(0, \sigma ^2)\).

The \(\SigmaNoisyLoad \) process:
Iteration: At each step \(t = 0, 1, 2, \ldots \),

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

  • • Compute the load estimates \(\tilde {x}_{i_1}^t = x_{i_1}^{t} + \mathcal {N}(0, \sigma ^2)\) and \(\tilde {x}_{i_2}^t = x_{i_2}^{t} + \mathcal {N}(0, \sigma ^2)\).

  • • Allocate to bin \(i \in \{ i_1, i_2 \}\) satisfying \(\tilde {x}_i^t = \min \{ \tilde {x}_{i_1}^t, \tilde {x}_{i_2}^t \}\).

Your browser does not support the HTML5 canvas tag.

.