Information Technology Reference
In-Depth Information
x 2
x 2 + 2 x 1
x 4 + x 2
x 4 +4 x 2 x 1
x 4 + x 2 + x 1
x 4 + x 2 −2 x 1
x 1
x 1
x 3 +10 x 1
x 3 x 1
x 3 +3 x 1
T 1
T 2
T 3
Fig. 7.1 Nodes and paths in C = T 1 + T 2 + T 3 +···
Theorem 3.2 ( Certificate for a non-identity) Let I be an ideal generated by some
multiplication terms. Let C
= i ↆ[ k ]
T i be a depth- 3 circuit that is nonzero modulo
3
I . Then
i
ↆ{
0
,...,
k
1
}
such that C
mod I has a path p satisfying: C
[
i
]
for some α ↆ F .
α ·
T i + 1 ≥∅
0
(
mod I
+∩
p
)
The proof of this theorem involves an extension of Chinese remaindering to ideals
that are generated by multiplication terms. Once we have this structural result about
depth-3, observe that we would be done if we could somehow ensure T i + 1
(in
our application I is zero). How do we preserve this ideal non-membership under a
cheap map?
Notice that the rank of the set S 0 of linear polynomials that divide the nodes in
the path p is
/ ↆ∩
p
k (since path length is below k ). Moreover, T i + 1 factors into at most d
linear polynomials, denote the set by S 1 . So if we apply a map that preserves the rank
of each of the d sets S 0 ⃒{ ʲ }
<
S 1 , then, intuitively, the ideal non-membership
(
S 0 ⃒{ ʲ } )
should be preserved. As rk
k we can employ the previously discussed
dnk 2 ). This idea could be easily turned into a
map
β
(over a field satisfying
|F| >
proof; details are in [ SS12 ].
Finally, what we have achieved is the construction of amap
,
that reduces the variables of C from n to k and preserves nonzeroness. Once this is
done, the poly
β , in time poly
(
dnk
)
nd k
(
)
blackbox PIT follows from the brute-force hitting-set.
7.3.2 Depth
3 Results
Looking at the success of bounded fanin depth-3 one wonders about the analogous
depth-4 model:
=
f i , j ,
C
where f i , j are sparse polynomials.
(7.1)
i
ↆ[
k
]
j
ↆ[
d
]
3 We mean C [ i ] := j ↆ[ i ] T j .
 
Search WWH ::




Custom Search