To define a unique, satisfying expected value from chosen sequences of bounded functions converging to an everywhere surjective function, we must carefully interpret the convergence, surjectivity, and the expectation framework. Here’s a step-by-step conceptual breakdown:
1. Function Setting
Let {fn}\{f_n\} be a sequence of bounded functions fn:[0,1]→Rf_n: [0,1] \to \mathbb{R}, converging pointwise (or in some sense) to a function ff, where ff is everywhere surjective (i.e., for every open interval (a,b)⊂R(a, b) \subset \mathbb{R}, there exists x∈[0,1]x \in [0,1] such that f(x)∈(a,b)f(x) \in (a,b)).
2. Expected Value Framework
We treat the expected value E[fn]\mathbb{E}[f_n] as:
since each fnf_n is bounded and Lebesgue integrable over [0,1][0,1].
3. Challenge with the Limit Function
The limiting function ff being everywhere surjective implies it's highly discontinuous — possibly not Lebesgue integrable, or even not measurable. Therefore, ∫01f(x)dx\int_0^1 f(x) dx might not exist or be meaningful.
4. Defining Expected Value via Limits
To define a unique, satisfying expected value, one strategy is:
This assumes the limit exists and is independent of the approximating sequence {fn}\{f_n\}.
But here's the catch: different sequences {fn}\{f_n\} converging to the same ff might yield different expectations, especially when ff is pathological.
5. Resolving the Ambiguity
To ensure uniqueness and satisfaction:
Restrict the mode of convergence: Use uniform convergence or convergence in L1L^1, not just pointwise.
Impose additional structure: e.g., if the fnf_n are continuous and uniformly bounded, and converge in L1L^1, then:limn→∞E[fn]=∫01f(x)dx\lim_{n \to \infty} \mathbb{E}[f_n] = \int_0^1 f(x) dxwhere f∈L1f \in L^1 (if possible).
6. If the Limit Function Is Not Integrable
If f∉L1f \notin L^1, or if the integral doesn’t exist in a standard sense, we can instead define the expected value as a limit of expectations of approximating functions:
provided the limit exists and is independent of the chosen sequence {fn}\{f_n\}. This approach avoids the direct integration of a possibly non-measurable function and relies on carefully controlled convergence and function properties.