Answers are grounded in this paper and cite their sources.
Let the random (stock market) vector <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">X ≥ 0</tex> be drawn according to a known distribution function <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F(x), x ∈ Rᵐ</tex> . A log-optimal portfolio <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b^{}</tex> is any portfolio <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</tex> achieving maximal expected <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">log</tex> return <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W^{}=b E ln bᵗX</tex> , where the supremum is over the simplex <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b ≥ 0, ∑ᵢ₌₁ᵐ bᵢ = 1</tex> . An algorithm is presented for finding <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b^{}</tex> . The algorithm consists of replacing the portfolio <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</tex> by the expected portfolio <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b', bᵢ' = E(bᵢXᵢ/bᵗX)</tex> , corresponding to the expected proportion of holdings in each stock after one market period. The improvement in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W(b)</tex> after each iteration is lower-bounded by the Kullback-Leibler information number <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D(b'\|b)</tex> between the current and updated portfolios. Thus the algorithm monotonically improves the return <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</tex> . An upper bound on <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W^{}</tex> is given in terms of the current portfolio and the gradient, and the convergence of the algorithm is established.
No takes yet. Share an insight, caveat, or question.
Thomas M. Cover (1984) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: