Document single output (#53)

* Document single output

* Better titles

docs/src/faq.md

@@ -1,11 +1,24 @@
 # Frequently Asked Questions
 
+## Supported autodiff backends
+
+- Forward mode: [ForwardDiff.jl](https://github.com/JuliaDiff/ForwardDiff.jl)
+- Reverse mode: all the packages compatible with [ChainRules.jl](https://github.com/JuliaDiff/ChainRules.jl)
+
+In the future, we would like to add [Enzyme.jl](https://github.com/EnzymeAD/Enzyme.jl) support.
+
 ## Higher-dimensional arrays
 
-For simplicity, the examples only display functions that work on vectors.
-However, arbitrary array sizes are supported.
+For simplicity, our examples only display functions that eat and spit out vectors.
+However, arbitrary array shapes are supported, as long as the forward _and_ conditions callables return similar arrays.
 Beware however, sparse arrays will be densified in the differentiation process.
 
+## Scalar input / output
+
+Functions that eat or spit out a single number are not supported.
+The forward _and_ conditions callables need arrays: for example, instead of returning `value` you should return `[value]` (a 1-element `Vector`). 
+Consider using an `SVector` from [StaticArrays.jl](https://github.com/JuliaArrays/StaticArrays.jl) if you seek increased performance.
+
 ## Multiple inputs / outputs
 
 In this package, implicit functions can only take a single input array `x` and output a single output array `y` (plus the additional info `z`).
@@ -26,7 +39,7 @@ f(x::ComponentVector) = f(x.a, x.b)
 
 The same trick works for multiple outputs.
 
-## Constrained optimization modeling
+## Modeling constrained optimization problems
 
 To express constrained optimization problems as implicit functions, you might need differentiable projections or proximal operators to write the optimality conditions.
 See [_Efficient and modular implicit differentiation_](https://arxiv.org/abs/2105.15183) for precise formulations.
@@ -37,10 +50,3 @@ In case these operators are too complicated to code them yourself, here are a fe
 - [ProximalOperators.jl](https://github.com/JuliaFirstOrder/ProximalOperators.jl)
 
 An alternative is differentiating through the KKT conditions, which is exactly what [DiffOpt.jl](https://github.com/jump-dev/DiffOpt.jl) does for JuMP models.
-
-## Which autodiff backends are supported?
-
-- Forward mode: ForwardDiff.jl
-- Reverse mode: all the packages compatible with [ChainRules.jl](https://github.com/JuliaDiff/ChainRules.jl)
-
-In the future, we would like to add [Enzyme.jl](https://github.com/EnzymeAD/Enzyme.jl) support.