Browsing by Person "Ahmadi, Hamed"

Now showing 1 - 3 of 3

Bi-objective optimization of nutrient intake and performance of broiler chickens using Gaussian process regression and genetic algorithm
(2023) Ahmadi, Hamed; Rodehutscord, Markus; Siegert, Wolfgang
This study investigated whether quantifying the trade-off between the maxima of two response traits increases the accuracy of diet formulation. To achieve this, average daily weight gain (ADG) and gain:feed ratio (G:F) responses of 7–21-day-old broiler chickens to the dietary supply of three nutrients (intake of digestible glycine equivalents, digestible threonine, and total choline) were modeled using a newly developed hybrid machine learning-based method of Gaussian process regression and genetic algorithm. The dataset comprised 90 data lines. Model-fit-criteria indicated a high model adjustment and no prediction bias of the models. The bi-objective optimization scenarios through Pareto front revealed the trade-off between maximized ADG and maximized G:F and provided information on the needed input of the three nutrients that interact with each other to achieve the trade-off scenarios. The trade-off scenarios followed a nonlinear pattern. This indicated that choosing target values intermediate to maximized ADG and G:F after single-objective optimization is less accurate than feed formulation after quantifying the trade-off. In conclusion, knowledge of the trade-off between maximized ADG and maximized G:F and the needed nutrient inputs will help feed formulators to optimize their feed with a more holistic approach.
Nutrient–response modeling with a single and interpretable artificial neuron
(2025) Rodehutscord, Markus; Ahmadi, Hamed
Precise estimation of nutrient requirements and utilization efficiency is fundamental to nutritional sciences, yet it is mainly performed using classical nonlinear regression models. These models are interpretable but require careful selection of the functional form and initial parameter values. Flexible machine learning (ML) methods are seemingly disliked due to their perceived “black box” nature, which can obscure biological insight. A minimal and interpretable ML framework addresses this gap in nutrient–response modeling. The proposed approach uses a single artificial neuron with a hyperbolic tangent activation. Mathematically, this resembles a four-parameter sigmoidal function but with greater flexibility and distinct parameter definitions, allowing capture of the monotonic, saturating dynamics typical of essential nutrient responses. The method is enhanced with modern ML best practices, including data augmentation, Bayesian regularization, and bootstrap resampling, providing robust, uncertainty-quantified estimates of key nutritional metrics—such as asymptotic response, inflection point, and nutrient requirements—even from small datasets. Evaluations across 12 diverse datasets from poultry and fish studies, including amino acids and phosphorus, demonstrated that the single artificial neuron matches or exceeds the performance of classical models while providing full analytical transparency. The framework is implemented as a no-code graphical application, ‘NutriCurvist’, offering an easy-to-use alternative tool for nutrient-response modeling to support data-driven, precision nutrition.
Systematic selection of best performing mathematical models for in vitro gas production using machine learning across diverse feeds
(2025) Ahmadi, Hamed; Titze, Natascha; Wild, Katharina; Rodehutscord, Markus
In vitro gas production (GP) is commonly used to evaluate ruminant feed, yet its accurate interpretation requires robust mathematical modeling. This study systematically explores a wide array of nonlinear models to explain GP dynamics across various feed types, addressing the question: how can efficient and versatile models that accurately represent GP profiles be identified? We hypothesized that distinct feed types exhibit unique GP characteristics, effectively captured by specific models, and that statistical and machine learning methodologies can streamline model selection. Utilizing a comprehensive dataset derived from 849 unique GP profiles across concentrate feed categories—including cereal and leguminous grains and processed protein feeds—21 candidate models were rigorously evaluated based on their goodness-of-fit metrics, with a particular emphasis on Bayesian Information Criterion (BIC) for model selection. A group of three models—namely Burr XII, Inverse paralogistic, and Log-logistic—consistently emerged as top performers, demonstrating high generalizability and predictive power across feed types. Notably, our analysis indicated that model type significantly influenced GP predictions, surpassing the impact of feed type characteristics. This research establishes a decision-making framework for model selection and sets the stage for further investigations linking in vitro GP parameters to in vivo digestibility, ultimately enhancing ruminant nutrition strategies.