Learning from the Past: Supplementary Exercise on Memory, Persistence and Explainable Outreach

Jonas, M. ORCID: https://orcid.org/0000-0003-1269-4145 & Zebrowski, P. ORCID: https://orcid.org/0000-0001-5283-8049 (2017). Learning from the Past: Supplementary Exercise on Memory, Persistence and Explainable Outreach. IIASA Working Paper. IIASA, Laxenburg, Austria: WP-17-016

[thumbnail of WP-17-016.pdf]
WP-17-016.pdf - Published Version
Available under License Creative Commons Attribution.

Download (826kB) | Preview


This Working Paper [WP] supports WP-17-015.
Toward Handling Uncertainty in Prognostic Scenarios: Advanced Learning from the Past
by Żebrowski, Jonas & Jarnicka (2017). Their WP (ZJJ WP hereafter) constitutes the main report summarizing the outcome of a one-year project (bearing the same title) under the Earth Systems Sciences [ESS] Research Program of the Austrian Academy of Sciences [OeAW].
The WP focuses on systems with memory, typical in Earth system sciences. Memory allows referring to how strongly a system’s past can influence its near-term future (paraphrased credibility of expectations about a system’s future behavior in the ZJJ WP) by virtue of its persistence. We consider memory an intrinsic property of the system, retrospective in nature; and persistence a consequential (observable) feature of memory, prospective in nature. We delineate the system’s near-term future by means of (what we call) its explainable outreach [EO].
This approach to determine the EO of a system complements the approach taken in the ZJJ WP. The WP makes use of a simple synthetic data (time) series example—our control—which we equip, step by step, with realistic physical features such as memory and noise, while exploring the system’s persistence and deriving its EO. The prime intention of the WP is to better understand memory and persistence and to consolidate our systems thinking. Therefore, during this explorative state, systemic insight is valued more than mathematical rigor. The example is geared to making the concept of EOs applicable. However, we discuss how consequential it is, where it underperforms, and the questions it provokes. From our example we conclude that memory allows defining a system’s explainable outreach, above and beyond the numerical set up given here. It seems that, even if we know only the temporal extent of memory, a system’s EO can be determined. This is promising because it appears possible to determine the temporal extent of memory in the presence of great noise, not exactly but approximately. However, even with complete knowledge of how memory evolves over time, we are confronted with the challenge of reconstructing best-fit regressions that separate memory and noise—a challenge that we leave for the future.

Item Type: Monograph (IIASA Working Paper)
Research Programs: Advanced Systems Analysis (ASA)
Related URLs:
Depositing User: Luke Kirwan
Date Deposited: 21 Sep 2017 06:31
Last Modified: 27 Aug 2021 17:29
URI: https://pure.iiasa.ac.at/14833

Actions (login required)

View Item View Item