## Abstract

We give a deterministic polynomial time algorithm for polynomial identity testing in the following two cases: 1. Non Commutative Arithmetic Formulas: The algorithm gets as an input an arithmetic formula in the non-commuting variables x_{1},...,x_{n} and determines whether or not the output of the formula is identically 0 (as a formal expression). 2. Pure Arithmetic Circuits: The algorithm gets as an input a pure arithmetic circuit (as defined by Nisan and Wigderson [3]) in the variables x_{1},...,x_{n} and determines whether or not the output of the circuit is identically 0 (as a formal expression). We also give a deterministic polynomial time identity testing algorithm for non commutative algebraic branching programs as defined by Nisan [2]. One application is a deterministic polynomial time identity testing for multilinear arithmetic circuits of depth 3. Finally, we observe an exponential lower bound for the size of pure arithmetic circuits for the permanent and for the determinant. (Only lower bounds for the depth of pure circuits were previously known [3]).

Original language | English |
---|---|

Pages (from-to) | 215-222 |

Number of pages | 8 |

Journal | Proceedings of the Annual IEEE Conference on Computational Complexity |

Volume | 19 |

State | Published - 2004 |

Externally published | Yes |

Event | Proceedings - 19th IEEE Annual Conference on Computational Complexity - Amherst, MA, United States Duration: 21 Jun 2004 → 24 Jun 2004 |