Package jline.api.cache

Class Cache_miss_rmf

java.lang.Object
jline.api.cache.Cache_miss_rmf
public final class Cache_miss_rmf extends Object
Refined mean-field (1/N-accurate) miss rates for RANDOM(m) caches.

Java port of the MATLAB cache_miss_rmf.m / Python cache_miss_rmf. Mirrors the Cache_miss_fpi contract: the popularity is recovered from the per-user per-item arrival rates (list-0 column), the mean-field fixed point is refined with its 1/N correction via CacheRMF, and the per-item miss probability is the steady-state occupancy of list 0.

Reference: N. Gast, "Expected Values Estimated via Mean-Field Approximation are 1/N-Accurate", Proc. ACM Meas. Anal. Comput. Syst., 2017.

  • Method Details

    • cache_miss_rmf

      public static CacheMissFpiResult cache_miss_rmf(Matrix gamma, Matrix m, Matrix[] lambdaCache)
      Compute cache miss rates using the refined mean-field method.
      Parameters:
      gamma - item access factors (accepted for interface parity with Cache_miss_fpi; used only for sizing fallbacks).
      m - cache capacity vector (h,).
      lambdaCache - per-user per-item per-list arrival rates: lambdaCache[v] is an (n x (h+1)) matrix; column 0 carries the request rates.
      Returns:
      CacheMissFpiResult with global miss rate M, per-user MU, per-item MI, and per-item miss probabilities pi0.

    • cache_miss_rmf_tran

      public static CacheMissRmfTranResult cache_miss_rmf_tran(Matrix gamma, Matrix m, Matrix[] lambdaCache, double time, int nPoints, double[] xinit)
      Transient refined mean-field cache trajectory. Integrates the same plain mean-field drift that cache_miss_rmf(jline.util.matrix.Matrix, jline.util.matrix.Matrix, jline.util.matrix.Matrix[]) drives to steady state over a finite window, from a supplied (or default) initial occupancy.

      Mirrors the MATLAB cache_miss_rmf.m tspan path (order-0, no 1/N correction in the transient).

      Parameters:
      gamma - item access factors (accepted for interface parity; used only for sizing).
      m - cache capacity vector (h,).
      lambdaCache - per-user per-item per-list arrival rates.
      time - end time of the transient window (start is 0).
      nPoints - number of uniform grid points (>= 2).
      xinit - initial occupancy (modelDimension,), or null for the default first-m-in-list initial state.
      Returns:
      CacheMissRmfTranResult with the time grid, per-item list-0 occupancy, per-user miss-rate trajectory, and full occupancy trajectory.