講座題目：Intelligent Annuities: Adaptive Retirement Planning with Deep RL（智能年金：基于深度強(qiáng)化學(xué)習(xí)的自適應(yīng)退休規(guī)劃）
主講人：金卓 澳大利亞麥考瑞大學(xué)商學(xué)院 教授
講座時(shí)間：2025年12月15日14:00
講座地點(diǎn)：學(xué)院229
講座主題摘要：
Abstract: This talk introduces a deep reinforcement learning framework specifically designed for the annuity market. We present an approach that dynamically optimizes portfolios combining variable annuities and traditional assets by directly learning from interactions with a simulated environment. The model integrates key annuity market uncertainties—like longevity risk and product-specific guarantees—into its decision process. By leveraging neural networks, the algorithm learns adaptive strategies to allocate between annuities and financial assets in response to evolving personal and market conditions. Our results demonstrate that this method provides a robust, data-driven, and personalized strategy for managing retirement income in complex annuity markets.
本報(bào)告介紹了一個(gè)專(zhuān)為年金市場(chǎng)設(shè)計(jì)的深度強(qiáng)化學(xué)習(xí)框架。研究者提出一種方法，通過(guò)直接從與模擬環(huán)境的交互中學(xué)習(xí)，動(dòng)態(tài)優(yōu)化包含變額年金與傳統(tǒng)資產(chǎn)的投資組合。該模型將年金市場(chǎng)的核心不確定性——?如長(zhǎng)壽風(fēng)險(xiǎn)及產(chǎn)品特定擔(dān)保條款?——?融入決策過(guò)程。借助神經(jīng)網(wǎng)絡(luò)，該算法能夠?qū)W習(xí)自適應(yīng)策略，根據(jù)個(gè)人狀況與市場(chǎng)環(huán)境的動(dòng)態(tài)變化，在年金與金融資產(chǎn)之間進(jìn)行配置。研究結(jié)果表明，該方法在復(fù)雜的年金市場(chǎng)中，為退休收入管理提供了一套穩(wěn)健、數(shù)據(jù)驅(qū)動(dòng)且個(gè)性化的解決方案。
主講人學(xué)術(shù)簡(jiǎn)介：
金卓，澳大利亞麥考瑞大學(xué)(Macquarie University)商學(xué)院教授，研究方向包括最優(yōu)控制論在精算中的應(yīng)用，數(shù)理金融，金融科技，機(jī)器學(xué)習(xí)與金融交叉。在Insurance Mathematics and Economics, European Journal of Operational Research, Journal of Risk and Insurance, SIAM Journal on Control and Optimization, Automatica等期刊發(fā)表論文60余篇。他是北美精算學(xué)會(huì)（Society of Actuaries）成員。
