- It is impossible to deduce all the terms from the original recursive formula.
r1 = Block[{a = 1},
F[a] - Sum[(-1)^a Binomial[a, k] Log[2]^(a - k) F[k], {k, 0, a}] ==
0 // FullSimplify]
r2 = Block[{a = 2},
F[a] - Sum[(-1)^a Binomial[a, k] Log[2]^(a - k) F[k], {k, 0, a}] ==
0 // FullSimplify]
r3 = Block[{a = 3},
F[a] - Sum[(-1)^a Binomial[a, k] Log[2]^(a - k) F[k], {k, 0, a}] ==
0 // FullSimplify]
r4 = Block[{a = 4},
F[a] - Sum[(-1)^a Binomial[a, k] Log[2]^(a - k) F[k], {k, 0, a}] ==
0 // FullSimplify]

Observe the first 4 recursive formulas. We can see that we can deduce F[1] from F[0] by r1 or r2( actually r1 is equivalent to r2).
And we can also deduce F[3] from F[0],F[1],F[2] by r3 or r4, but we still cann't deduce F[2]. (BTW, under the condition of r1, the r3 and r4 are equivalent)
- The expression of
f[2] as below indicate that f[2] it is unlikely can be deduce from f[0] and f[1].
f[a_] := Integrate[(Log[x/(1 - x)])^a/(1 + x), {x, 0, 1}];
{f[0], f[1], f[2]}
% // N
Table[NIntegrate[(Log[x/(1 - x)])^a/(1 + x), {x, 0, 1}], {a, {0, 1,
2}}]

F[a], where $a$ is even from these equations. For instance you get for $a=2$:F[2]==F[2] - 2 F[1] Log[2] + F[0] Log[2]^2. Your question has no solution. $\endgroup$F[2]orF[3]? Also it isBinomialnotbinomial$\endgroup$