-
Multi-Objective Bayesian Optimization with Active Preference Learning
Authors:
Ryota Ozaki,
Kazuki Ishikawa,
Youhei Kanzaki,
Shinya Suzuki,
Shion Takeno,
Ichiro Takeuchi,
Masayuki Karasuyama
Abstract:
There are a lot of real-world black-box optimization problems that need to optimize multiple criteria simultaneously. However, in a multi-objective optimization (MOO) problem, identifying the whole Pareto front requires the prohibitive search cost, while in many practical scenarios, the decision maker (DM) only needs a specific solution among the set of the Pareto optimal solutions. We propose a B…
▽ More
There are a lot of real-world black-box optimization problems that need to optimize multiple criteria simultaneously. However, in a multi-objective optimization (MOO) problem, identifying the whole Pareto front requires the prohibitive search cost, while in many practical scenarios, the decision maker (DM) only needs a specific solution among the set of the Pareto optimal solutions. We propose a Bayesian optimization (BO) approach to identifying the most preferred solution in the MOO with expensive objective functions, in which a Bayesian preference model of the DM is adaptively estimated by an interactive manner based on the two types of supervisions called the pairwise preference and improvement request. To explore the most preferred solution, we define an acquisition function in which the uncertainty both in the objective functions and the DM preference is incorporated. Further, to minimize the interaction cost with the DM, we also propose an active learning strategy for the preference estimation. We empirically demonstrate the effectiveness of our proposed method through the benchmark function optimization and the hyper-parameter optimization problems for machine learning models.
△ Less
Submitted 22 November, 2023;
originally announced November 2023.
-
Transforming U.S. agriculture with crushed rock for CO$_2$ sequestration and increased production
Authors:
David J. Beerling,
Euripides P. Kantzas,
Maria Val Martin,
Mark R. Lomas,
Lyla L. Taylor,
Shuang Zhang,
Yoshiki Kanzaki,
Christopher T. Reinhard,
Noah J. Planavsky,
Rafael M. Eufrasio,
Phil Renforth,
Jean-Francois Mecure,
Hector Pollitt,
Philip B. Holden,
Neil R. Edwards,
Lenny Koh,
Dimitar Z. Epihov,
Adam Wolf,
James E. Hansen,
Nick F. Pidgeon,
Steven A. Banwart
Abstract:
Enhanced weathering (EW) is a promising modification to current agricultural practices that uses crushed silicate rocks to drive carbon dioxide removal (CDR). If widely adopted on farmlands, it could help achieve net-zero or negative emissions by 2050. We report detailed state-level analysis indicating EW deployed on agricultural land could sequester 0.23-0.38 Gt CO$_2$ yr$^{-1}$ and meet 36-60 %…
▽ More
Enhanced weathering (EW) is a promising modification to current agricultural practices that uses crushed silicate rocks to drive carbon dioxide removal (CDR). If widely adopted on farmlands, it could help achieve net-zero or negative emissions by 2050. We report detailed state-level analysis indicating EW deployed on agricultural land could sequester 0.23-0.38 Gt CO$_2$ yr$^{-1}$ and meet 36-60 % of U.S. technological CDR goals. Average CDR costs vary between state, being highest in the first decades before declining to a range of $\sim\$$100-150 tCO$_2{}^{-1}$ by 2050, including for three states (Iowa, Illinois, and Indiana) that contribute most to total national CDR. We identify multiple electoral swing states as being essential for scaling EW that are also key beneficiaries of the practice, indicating the need for strong bipartisan support of this technology. Assessment the geochemical capacity of rivers and oceans to carry dissolved EW products from soil drainage suggests EW provides secure long-term CO$_2$ removal on intergenerational time scales. We additionally forecast mitigation of ground-level ozone increases expected with future climate change, as an indirect benefit of EW, and consequent avoidance of yield reductions. Our assessment supports EW as a practical innovation for leveraging agriculture to enable positive action on climate change with adherence to federal environmental justice priorities. However, implementing a stage-gating framework as upscaling proceeds to safeguard against environmental and biodiversity concerns will be essential.
△ Less
Submitted 8 August, 2023;
originally announced August 2023.
-
Oceanic and atmospheric methane cycling in the cGENIE Earth system model
Authors:
Christopher T. Reinhard,
Stephanie L. Olson,
Sandra Kirtland Turner,
Cecily Palike,
Yoshiki Kanzaki,
Andy Ridgwell
Abstract:
The methane cycle is a key component of the Earth system that links planetary climate, biological metabolism, and the global biogeochemical cycles of carbon, oxygen, sulfur, and hydrogen. However, currently lacking is a numerical model capable of simulating a diversity of environments in the ocean where methane can be produced and destroyed, and with the flexibility to be able to explore not only…
▽ More
The methane cycle is a key component of the Earth system that links planetary climate, biological metabolism, and the global biogeochemical cycles of carbon, oxygen, sulfur, and hydrogen. However, currently lacking is a numerical model capable of simulating a diversity of environments in the ocean where methane can be produced and destroyed, and with the flexibility to be able to explore not only relatively recent perturbations to Earth's methane cycle but also to probe methane cycling and associated climate impacts under the reducing conditions characteristic of most of Earth history and likely widespread on other Earth-like planets. Here, we present an expansion of the ocean-atmosphere methane cycle in the intermediate-complexity Earth system model cGENIE, including parameterized atmospheric photochemistry and schemes for microbial methanogenesis, aerobic methanotrophy, and anaerobic oxidation of methane. We describe the model framework, compare model parameterizations against modern observations, and illustrate the flexibility of the model through a series of example simulations. Though we make no attempt to rigorously tune default model parameters, we find that simulated atmospheric methane levels and marine dissolved methane distributions are generally in good agreement with empirical constraints for the modern and recent Earth. Finally, we illustrate the model's utility in understanding the time-dependent behavior of the methane cycle resulting from transient carbon injection into the atmosphere, and present model ensembles that examine the effects of oceanic chemistry and the thermodynamics of microbial metabolism on steady-state atmospheric methane abundance.
△ Less
Submitted 29 July, 2020;
originally announced July 2020.
