OnTrialSetandUncorrectableErrorsfortheFirst-Order.pptVIP

On Trial Set and Uncorrectable Errors for the First-Order Reed-Muller Codes Kenji Yasunaga Toru Fujiwara Osaka University 2007 Hawaii and SITA Joint Conference on Information Theory (HISC2007), Hawaii, USA, May 29-31, 2007 * Trial set Trial set T for a binary linear code C is a subset of C that meets some property (describe later) introduced by [Helleseth, Kl?ve, Levenshtein, 2005] used for Maximum Likelihood Decoding (MLD) upper bounding #(uncorrectable errors) by MLD Smaller trial set is desirable ? How large is the size of minimum trial set Tmin ? * Main results For binary linear codes Give upper/lower bounds on |Tmin| For the first-order Reed-Muller codes RMm Determine |Tmin| Determine #(minimal uncorrectable errors) * Contents Notations Background Coset partitioning and Syndrome decoding Monotone structure of Errors Trial set, Minimal uncorrectable errors, Larger half Main Results Upper/lower bounds on |Tmin| for linear codes |Tmin| for RMm #(minimal uncorrectable errors) in RMm * Notations Support set of x; S(x) := { i : xi≠0 } x covers y; x ? y ? S(x) ? S(y) x is lexicographically smaller than y; x y ? ∥x∥ ∥y∥  or ∥x∥ = ∥y∥ and v(x) v(y) Hamming weight of x; ∥x∥:= |S(x)| Numerical value of x; v(x) := Σxi2n-i Example.  000 001 010 100 011 101 110 111 * Coset partitioning and Syndrome decoding Coset partitioning Fn = Ci, Ci∩Cj ＝Φ for i≠j, Ci := {vi+c : c∈C} : a coset vi ∈Fn : a coset leader Syndrome decoding y∈Fn : a received vector Output y+vi if y∈Ci Coset leaders are correctable errors. If vi has minimum weight in Ci, it performs MLD. ∪ i=1 2n-k If each coset leader is lexicographically smallest in its coset, errors have monotone structure. * Monotone structure of errors E0(C) := the set of coset leaders (= Correctable errors) E1(C) := Fn?E0(C) (= Uncorrectable errors) Monotone structure of errors Suppose x∈E0(C), y∈E1(C). Al