A Bayesian sequential updating approach to predict phenology of silage maize

Crop models are tools used for predicting year to year crop development on field to regional scales. However, robust predictions are hampered by factors such as uncertainty in crop model parameters and in the data used for calibration. Bayesian calibration allows for the estimation of model parameters and quantification of uncertainties, with the consideration of prior information. In this study, we used a Bayesian sequential updating (BSU) approach to progressively incorporate additional data at a yearly time-step to calibrate a phenology model (SPASS) while analysing changes in parameter uncertainty and prediction quality. We used field measurements of silage maize grown between 2010 and 2016 in the regions of Kraichgau and Swabian Alb in southwestern Germany. Parameter uncertainty and model prediction errors were expected to progressively reduce to a final, irreducible value. Parameter uncertainty reduced as expected with the sequential updates. For two sequences using synthetic data, one in which the model was able to accurately simulate the observations, and the other in which a single cultivar was grown under the same environmental conditions, prediction error mostly reduced. However, in the true sequences that followed the actual chronological order of cultivation by the farmers in the two regions, prediction error increased when the calibration data was not representative of the validation data. This could be explained by differences in ripening group and temperature conditions during vegetative growth. With implications for manual and automatic data streams and model updating, our study highlights that the success of Bayesian methods for predictions depends on a comprehensive understanding of inherent structure in the observation data and model limitations.

Change Matters: Medication Change Prediction with Recurrent Residual Networks

Deep learning is revolutionizing predictive healthcare, including recommending medications to patients with complex health conditions. Existing approaches focus on predicting all medications for the current visit, which often overlaps with medications from previous visits. A more clinically relevant task is to identify medication changes. In this paper, we propose a new recurrent residual networks, named MICRON, for medication change prediction. MICRON takes the changes in patient health records as input and learns to update a hid- den medication vector and the medication set recurrently with a reconstruction design. The medication vector is like the memory cell that encodes longitudinal information of medications. Unlike traditional methods that require the entire patient history for prediction, MICRON has a residual-based inference that allows for sequential updating based only on new patient features (e.g., new diagnoses in the recent visit), which is efficient. We evaluated MICRON on real inpatient and outpatient datasets. MICRON achieves 3.5% and 7.8% relative improvements over the best baseline in F1 score, respectively. MICRON also requires fewer parameters, which significantly reduces the training time to 38.3s per epoch with 1.5× speed-up.

Improvement of Side Resistance Prediction for Pile Foundation Using Construction Information

Problems such as the inclination and settlement of buildings after construction completion have been reported in recent years. Quality control during construction and mieruka (“visualizing” the work quality) are increasingly demanded. In light of this, the present study addresses execution with a “rotary penetration steel pipe pile (RPS-pile)”, which is a system that is able to collect information in real time during the piling process. When a RPS-pile is placed, high push-in and pull-out bearing capacities are expected because the spiral blade at the pile tip resists at the bottom of the borehole against an external force. Additionally, continuous real-time data on the pile-head torque and the auger penetration depth per revolution are obtained, as these mechanical indicators are necessary for piling. This study aims at developing a method for the real-time confirmation of work quality at construction sites by utilizing construction information (information obtained during piling). Specifically, a method is proposed for the sequential updating of reliability in the estimation of the side friction acting on a drilled pile. In this method, real-time information on the pile-head torque and the auger penetration depth per revolution obtained during piling are used in addition to N-values that are observed in a standard penetration test conducted in advance as part of subsurface exploration.

Sequential Updating of Quantitative Requirements for Increased Flexibility in Robust Systems Design

In early development stages of complex systems, interacting subsystems (including components) are often designed simultaneously by distributed teams with limited information exchange. Distributed development becomes possible by assigning teams independent design goals expressed as quantitative requirements equipped with tolerances to provide flexibility for design: so-called solution-spaces are high-dimensional sets of permissible subsystem properties on which requirements on the system performance are satisfied. Edges of box-shaped solution spaces are permissible intervals serving as decoupled (mutually independent) requirements for subsystem design variables. Unfortunately, decoupling often leads to prohibitively small intervals. In so-called solution-compensation spaces, permissible intervals for early-decision variables are increased by a compensation mechanism using late-decision variables. This paper presents a multi-step development process where groups of design variables successively change role from early-decision to late-decision type in order to maximize flexibility. Applying this to a vehicle chassis design problem demonstrates the effectiveness of the approach.