An avoidance learning situation. A neural net model

A simple avoidance situation is considered in terms of a neural net learning model. Data for the control situation can be represented by an expression having three parameters which determine the initial and the steady state activities together with the transient aspects. The introduction of a learning parameter then allows one to calculate satisfactorily the results obtained in the experimental situation in which shock is applied. This research was supported in part by the United States Air Force through the Air Force Office of Scientific Research of the Air Research Development Command under Grant No. AF AFOSR 370-63 and in part by the United States Public Health Service Grant RCA GM K6 18,420.     

A neural model for conditioned avoidance learning

A mathematical model for learning of a conditioned avoidance behavior is presented. An identification of the net excitation of a neural model (Rashevsky, N., 1960.Mathematical Biophysics. Vol. II. New York: Dover Publications, Inc.) with the instantaneous probability of response is introduced and its usefulness in discussing block-trial learning performances in the conditioned avoidance situation is outlined for normal and brain-operated animals, using experimental data collected by the author. Later, the model is applied to consecutive trial learning and connection is made with the approach of H. D. Landahl (1964. “An Avoidance Learning Situation. A Neural Net Model.”Bull. Math. Biophysics,26, 83–89; and 1965, “A Neural Net Model for Escape Learning.”Bull. Math. Biophysics,27, Special Edition, 317–328) wherein lie further data with which the model can be compared.     

Reinforcement learning for a stochastic automaton modelling predation in stationary model-mimic environments

In this paper we propose a mathematical learning model for the feeding behaviour of a specialist predator operating in a random environment occupied by two types of prey, palatable mimics and unpalatable models, and a generalist predator with additional alternative prey at its disposal. A well known linear reinforcement learning algorithm and its special cases are considered for updating the probabilities of the two actions, eat prey or ignore prey. Each action elicits a probabilistic response from the environment that can be favorable or unfavourable. To assess the performance of the predator a payoff function is constructed that captures the energetic benefit from consuming acceptable prey, the energetic cost from consuming unacceptable prey, and lost benefit from ignoring acceptable prey. Conditions for an improving predator payoff are also explicitly formulated.     

Understanding predation risk and individual variation in risk avoidance for threatened boreal caribou

Predation risk is a driver of species’ distributions. Animals can increase risk avoidance in response to fluctuations in predation risk, but questions remain regarding individual variability and the capacity to respond to changes in spatial risk across human‐altered landscapes. In northeast British Columbia, Canada, boreal caribou populations declined as roads and seismic lines have increased, which are theorized to increase gray wolf predation. Our goal was to model risk and to evaluate individual variability and the development of risk perception by examining individual risk avoidance in response to reproductive status and age. We used locations from collared caribou and wolves to identify landscape features associated with the risk of a potential wolf‐caribou encounter and risk of being killed given an encounter. We built resource selection functions to estimate individual responses to risk. We used general linear regressions to evaluate individual risk and linear feature avoidance as a function of age and reproductive status (calf or no calf). Linear features increased the risk of encounter. Older caribou and caribou with calves demonstrated stronger avoidance of the risk of encounter and roads, but weaker avoidance in late summer to the risk of being killed relative to younger and calf‐less individuals. Mechanisms explaining the inverse relationships between the risk of encounter and risk of being killed are uncertain, but it is conceivable that caribou learn to avoid the risk of encounter and roads. Responses by females with vulnerable calves to the risk of encounter and risk of being killed might be explained by a trade‐off between these two risk types and a prioritization on the risk of encounter. Despite the capacity to alter their responses to risk, the global decline in Rangifer populations (caribou and wild reindeer) suggests these behaviors are insufficient to mitigate the impacts of anthropogenic disturbances.     

Plasticity or fixed adaptive traits? Strategies for predation avoidance in Rana arvalis tadpoles

Tadpoles of Rana arvalis originating from seven island populations were tested for responses to non-lethal predator presence. In general, tadpole growth was reduced and the relative tail depth was increased at predator presence. There was no effect of predator presence on the predicted size at metamorphosis. The differentiation rate, translating as length of the larval period, was lower at predator presence, but this seems to be merely an effect of the reduced growth. Although populations differed with respect to growth, relative tail length, relative tail depth, differentiation rate and predicted size at metamorphosis, no obvious differences were found in their responses to predator presence. Data on predator occurrences in the source ponds show that tadpoles originating from ponds with a high predation pressure have a higher differentiation rate, i.e. they will metamorphose at an earlier date than those from “safe” ponds (if raised under the same conditions). Moreover, they are also predicted to metamorphose at a smaller size, which is in accordance with theoretical models. Despite the fact that populations differed in growth, no correlation was found between growth and predation risk in the source ponds. Received: 16 March 1998 / Accepted: 18 July 1998     

Towards a general model for protein-substrate stereoselectivity

Protein-substrate interactions in enzymatic, neurological, and immunological systems are typically characterized by a high degree of stereoselectivity towards complex substrates. We propose a novel stereocenter-recognition (SR) model for stereoselectivity of proteins (or receptors in general) towards substrates that have multiple stereocenters, based on the topology of substrate stereocenters. The model provides the minimum number of substrate locations that need to enter into binding, nonbinding, or repulsive interactions with receptor sites, for stereoselectivity to occur. According to this model, a substrate location may interact with multiple receptor sites, or multiple substrate locations may interact with a single receptor site, but a stereoselective receptor has to offer, in the correct geometry, at least as many interactions as the required minimum number of substrate locations. The SR model predicts that stereoselectivity towards an acyclic substrate with N stereocenters distributed along a single chain requires interactions involving a minimum of N + 2 substrate locations, distributed over all stereocenters in the substrate, such that effectively three locations exist per stereocenter. Thus, enantioselective recognition of molecules with one chiral center requires a protein to interact with a minimum of three substrate locations, while stereoselectivity towards substrates with two or three stereocenters requires interactions with a minimum of four or five substrate locations, respectively, and so on. We demonstrate the general applicability of this model to protein-substrate interactions by interpreting several previous experimental observations.     

An inquiry-based learning model for an exercise physiology laboratory course

We developed an inquiry-based learning model to better stimulate undergraduate students' cognitive development of exercise physiology laboratory concepts. The course core is the two independent research projects that students, working in small groups, complete during the last 9 wk of the semester. Student groups develop their own research question and hypothesis, design the experiment, collect and analyze the data, and report their findings to the rest of the class using presentation software. To help with success of the research projects, students are taken through a series of guided-inquiry laboratory activities during the initial 6 wk of the semester to develop laboratory skills and an understanding of the scientific process. Observations of student behaviors reflected a high level of enthusiasm and engagement in laboratory activities. Surveys, journal entries, and interviews indicated that students felt empowered by having ownership in their projects, which may be the key reason for the success of this model.     

Analysis of a sterile insect release model with predation

A model for the sterile insect release method of pest control in which the target species is under predatory or parasitic regulation is analyzed. The equations are nondimensionalized and the rescaled parameters are interpreted. There are four types of equilibria, whose existence and stability depend on which of ten regions of parameter space contain the rescaled parameters, and in turn give minimal release rates to achieve eradication of the pest. In at least one region, Hopf bifurcation theory shows the existence of limit cycles, but they are found to be unstable. In addition, the optimal release rate to minimize a total cost functional for pest control by the sterile release method is studied. Both approaches show that when predation accounts for a large fraction of the natural deaths, the necessary release rate and total cost are higher than for weak predation. If the predators are removed without being replaced by any other source of mortality, the cost rises in all cases but rises much more dramatically for cases with strong predation. A definite danger of the sterile release method when some predatory control exists is that the predators are frequently driven extinct before the prey, so that the target species could explode to much higher levels and be more difficult to eradicate again after the sterile release is terminated.     

An asymptotic theory for model selection inference in general semiparametric problems

Hjort & Claeskens (2003) developed an asymptotic theoryfor model selection, model averaging and subsequent inferenceusing likelihood methods in parametric models, along with associatedconfidence statements. In this article, we consider a semiparametricversion of this problem, wherein the likelihood depends on parametersand an unknown function, and model selection/averaging is tobe applied to the parametric parts of the model. We show thatall the results of Hjort & Claeskens hold in the semiparametriccontext, if the Fisher information matrix for parametric modelsis replaced by the semiparametric information bound for semiparametricmodels, and if maximum likelihood estimators for parametricmodels are replaced by semiparametric efficient profile estimators.Our methods of proof employ Le Cam's contiguity lemmas, leadingto transparent results. The results also describe the behaviourof semiparametric model estimators when the parametric componentis misspecified, and also have implications for pointwise-consistentmodel selectors.     

Temporal learning of predation risk by embryonic amphibians

For prey species that rely on learning to recognize their predators, natural selection should favour individuals able to learn as early as possible. The earliest point at which individuals can gather information about the identity of their potential predators is during the embryonic stage. Indeed, recent experiments have demonstrated that amphibians can learn to recognize predators prior to hatching. Here, we conditioned woodfrog embryos to recognize predatory salamander cues either in the morning or in the evening, and subsequently exposed the two-week-old tadpoles to salamander cues either in the morning or in the evening, and recorded the intensity of their antipredator behaviour. The data indicate that amphibians learn to recognize potential predators while still in the egg, and also learn the temporal component of this information, which they use later in life, to adjust the intensity of their antipredator responses throughout the day.     

Aquatic plant shows flexible avoidance by escape from tuber predation by swans

Deeper burial of bulbs and tubers has been suggested as an escape against below-ground herbivory by vertebrates, but experimental evidence is lacking. As deep propagule burial can incur high costs of emergence after dormancy, burial depth may represent a trade-off between sprouting survival and herbivore avoidance. We tested whether burial depth of subterraneous tubers is a flexible trait in fennel pondweed (Potamogeton pectinatus), facing tuber predation by Bewick's swans (Cygnus columbianus bewickii) in shallow lakes in winter. In a four-year experiment involving eight exclosures, winter herbivory by swans and all vertebrate summer herbivory were excluded in a full-factorial design; we hence controlled for aboveground vertebrate herbivory in summer, possibly influencing tuber depth. Tuber depth was measured each September before swan arrival and each March before tuber sprouting. In accordance with our hypothesis, tuber depth in September decreased after excluding Bewick's swans in comparison to control plots. The summer exclosure showed an increase in tuber biomass and the number of shallow tubers, but not a significant effect on the mean burial depth of tuber mass. Our results suggest that a clonal plant like P. pectinatus can tune the tuber burial depth to predation pressure, either by phenotypic plasticity or genotype sorting, hence exhibiting flexible avoidance by escape. We suggest that a flexible propagule burial depth can be an effective herbivore avoidance strategy, which might be more widespread among tuber forming plant species than previously thought.     

Predator detection and avoidance by starlings under differing scenarios of predation risk

Practically all animals must find food while avoiding predators.An individual's perception of predation risk may depend on manyfactors, such as distance to refuge and group size, but it isunclear whether individuals respond to different factors ina similar manner. We tested whether flocks of foraging starlingsresponded in the same way to an increased perception of predationrisk by assessing three factors: (1) neighbor distances, (2)habitat obstruction, and (3) recent exposure to a predator.We found that in all three scenarios of increased risk, starlingsreduced their interscan intervals (food-searching bouts), whichincreased the frequency of their vigilance periods. We thenexamined how one of these factors, habitat obstruction, affectedescape speed by simulating an attack with a model predator.Starlings were slower to respond in visually obstructed habitats(long grass swards) and slower when they had their head downin obstructed habitats than when they had their head down inopen habitats. In addition, reaction times were quicker whenstarlings could employ their peripheral fields of vision. Ourresults demonstrate that different sources of increased riskcan generate similar behavioral responses within a species.The degree of visibility in the physical and social environmentaffects both the actual and perceived risk of predation.     

Projection of climate variables by general circulation and deep learning model for Lahore,Pakistan

Climate change significantly impacts the hydrological cycle and environment. The key parameters driving climate change for densely populated city of Lahore in Pakistan were studied. The projections of these parameters were evaluated using General Circulation Model (GCM) named Community Climate System Model (CCSM4) under two Representative Concentration Pathways (RCP 4.5 and 8.5) scenarios. The outputs of CCSM4 model were bias corrected using quantile mapping using historical data. Additionally, a deep learning model named long short-term memory (LSTM) was developed applying machine learning applications to forecast the climate parameters for the future. LSTM model with two LSTM layers including one fully connected layer was modeled for the projection of climate variables in the region. Total number of parameters were 9888, and the input and forecasted output length was kept as 24 sequential months without overlapping. The conventional projection methods of GCM were compared with LSTM outputs for bridging the gap. The LSTM model was found to be more effective and dependable in forecasting the climate with significant improvement in the statistical parameters for the region. The LSTM model can be applied for projections of climate in comparison to GCM with sufficient precision.