For a polynomial f,
a weighted sum-of-squares representation (SOS) has the form
$$f = \sum_{i\in [s]} c_i f_i^2$$
f
=
∑
i
∈
[
s
]
c
i
f
i
2
, where the weights$$c_i$$
c
i
are field elements.
The size of the representation is the number of monomials that appear across the $$f_i$$
f
i
's.
Its minimum across all such decompositions is called the support-sum S(f) of f.For a univariate polynomial f of degree d of full support,
a lower bound for the support-sum is $$S(f) \ge \sqrt d$$
S
(
f
)
≥
d
.
We show that the existence of an explicit univariate polynomial f
with support-sum just slightly larger than the lower bound, that is, $$S(f) \ge d^{0.5+\varepsilon}$$
S
(
f
)
≥
d
0.5
+
ε
,
for some $$\varepsilon > 0$$
ε
>
0
,
implies that $$\ne$$
≠
,
the major open problem in algebraic complexity.
In fact, our proof works for some subconstant functions $$\varepsilon(d) > 0$$
ε
(
d
)
>
0
as well.
We also consider the sum-of-cubes representation (SOC) of polynomials. We show that an explicit hard polynomial
implies both blackbox-PIT is in , and $$\neq$$
≠
.