Search

Recent research in machine learning pointed to the core problem of state-of-the-art models which impedes their widespread adoption in different domains. The models' inability to differentiate between noise and subtle, yet significant variation in data leads to their vulnerability to adversarial perturbations that cause wrong predictions with high confidence. The study is aimed at identifying whether the algorithms inspired by biological evolution may achieve better results in cases where brittle robustness properties are highly sensitive to the slight noise. To answer this question, we introduce the new robust gradient descent inspired by the stability and adaptability of biological systems to unknown and changing environments. The proposed optimization technique involves an open-ended adaptation process with regard to two hyperparameters inherited from the generalized Verhulst population growth equation. The hyperparameters increase robustness to adversarial noise by penalizing the degree to which hardly visible changes in gradients impact prediction. The empirical evidence on synthetic and experimental datasets confirmed the viability of the bio-inspired gradient descent and suggested promising directions for future research. The code used for computational experiments is provided in a repository at https://github.com/yukinoi/bio_gradient_descent.

Managing the disturbance of visitors due to crowding is an important management task in protected areas with high use levels. To achieve this, managers need to know how the use level affects the perceived disturbance due to crowding. Here we present a method to predict the level of disturbance as a function of use level measured by number of visitors. In contrast to the visual approach where subjects are asked to evaluate acceptability of use levels from manipulated images of scenery, our approach uses data gathered from actual experiences: actual (measured) use levels and concurrent on-site data on levels of disturbance experienced by visitors. Using the example of Nature Park Telašćica, we show how these data can be acquired with limited resources (a smart-phone and short, time-stamped questionnaires), and demonstrate the subsequent analysis and model fitting. The resulting model estimates the probability that a visitor experiencing a given use level will report certain level of disturbance. We suggest a way of using the probability density functions to define an inherent limit of acceptable disturbance (LAD) due to crowding; the LAD can also be set to a desired value by management. Regardless of the definition, LAD can be used to determine the maximum acceptable use level as dictated by crowding considerations. The method gives predictions consistent with previous literature and can be used even when data are collected at low use levels.