Turing learning: a metric-free approach to inferring behavior and its application to swarms

We propose Turing Learning, a novel system identification method for inferring the behavior of natural or artificial systems. Turing Learning simultaneously optimizes two populations of computer programs, one representing models of the behavior of the system under investigation, and the other representing classifiers. By observing the behavior of the system as well as the behaviors produced by the models, two sets of data samples are obtained. The classifiers are rewarded for discriminating between these two sets, that is, for correctly categorizing data samples as either genuine or counterfeit. Conversely, the models are rewarded for ‘tricking’ the classifiers into categorizing their data samples as genuine. Unlike other methods for system identification, Turing Learning does not require predefined metrics to quantify the difference between the system and its models. We present two case studies with swarms of simulated robots and prove that the underlying behaviors cannot be inferred by a metric-based system identification method. By contrast, Turing Learning infers the behaviors with high accuracy. It also produces a useful by-product—the classifiers—that can be used to detect abnormal behavior in the swarm. Moreover, we show that Turing Learning also successfully infers the behavior of physical robot swarms. The results show that collective behaviors can be directly inferred from motion trajectories of individuals in the swarm, which may have significant implications for the study of animal collectives. Furthermore, Turing Learning could prove useful whenever a behavior is not easily characterizable using metrics, making it suitable for a wide range of applications.     

A taxonomic revision of <Emphasis Type="Italic">Bursera</Emphasis> subgen. <Emphasis Type="Italic">Bursera</Emphasis> in the Greater Antilles and the Bahamas,including a new species from Cuba

A revision of Bursera in the Greater Antilles and the Bahamas has confirmed that Commiphora does not occur in the region and that all but one species in the region belong to Bursera subgen. Bursera. Here we describe a new species, Bursera yaterensis; B. nashii is synonymized with B. glauca and B. ovata is synonymized with B. trinitensis; and we return five species from Commiphora to Bursera. A dichotomous key is provided using mostly vegetative characters due to the frequent lack of adequate reproductive material and the relative uniformity of most floral and fruit characters.     

Scale-dependent bi-trophic interactions in a semi-arid savanna: how herbivores eliminate benefits of nutrient patchiness to plants

The scale of resource heterogeneity may influence how resources are locally partitioned between co-existing large and small organisms such as trees and grasses in savannas. Scale-related plant responses may, in turn, influence herbivore use of the vegetation. To examine these scale-dependent bi-trophic interactions, we varied fertilizer [(nitrogen (N)/phosphorus (P)/potassium (K)] applications to patches to create different scales of nutrient patchiness (patch size 2 × 2 m, 10 × 10 m, or whole-plot 50 × 50 m) in a large field experiment in intact African savanna. Within-patch fertilizer concentration and the total fertilizer load per plot were independently varied. We found that fertilization increased the leaf N and P concentrations of trees and grasses, resulting in elevated utilization by browsers and grazers. Herbivory off-take was particularly considerable at higher nutrient concentrations. Scale-dependent effects were weak. The net effect of fertilization and herbivory was that plants in fertilized areas tended to grow less and develop smaller rather than larger standing biomass compared to plants growing in areas that remained unfertilized. When all of these effects were considered together at the community (plot) level, herbivory completely eliminated the positive effects of fertilization on the plant community. While this was true for all scales of fertilization, grasses tended to profit more from coarse-grained fertilization and trees from fine-grained fertilization. We conclude that in herbivore-dominated communities, such as the African savanna, nutrient patchiness results in the herbivore community profiting rather more than the plant community, irrespective of the scale of patchiness. At the community level, the allometric scaling theory’s prediction of plant—and probably also animal—production does not hold or may even be reversed as a result of complex bi-trophic interactions.     

Short-term partitioning of Cd recently taken up between sunflowers organs (<Emphasis Type="Italic">Helianthus annuus</Emphasis>) at flowering and grain filling stages: effect of plant transpiration and allometry

Aims

This work concentrated on understanding the allocation of Cd recently taken up between the organs of sunflower at early and middle reproductive growth stages. The roles of transpiration and allometry were investigated.

Methods

Sunflowers were grown hydroponically in greenhouse, being exposed to low concentrations of Cd (pCd²⁺ = 11.03). At flower bud and grain filling stages, plants were exposed for three days to ¹¹¹Cd and at the same time, subjected or not to fans to increase the transpiration. The partitioning of ¹¹¹Cd between plant organs measured by high resolution ICP-MS was then modelled.

Results

Although the use of fans increased the plant water uptake and transpiration by about 20%, there were no significant effects on the partitioning of recent Cd. Most of the recent Cd was recovered in roots (60%) and only 2.8% were found in seeds (0.8% for the husk and 2.0% for the almonds). The sequestration of recent Cd in a plant organ was successfully explained by its biomass and except for leaves, by the biomass of other organs acting as competitive sinks.

Conclusions

This work proposes a modelling approach for the partitioning of the labelled Cd between plant organs in sunflower.

     

Linking root traits and competitive success in grassland species

Background and aims

Measures of phosphorus (P) in roots recovered from soil underestimate total P accumulation below-ground by crop species since they do not account for P in unrecovered (e.g., fine) root materials. ³³P-labelling of plant root systems may allow more accurate estimation of below-ground P input by plants.

Methods

Using a stem wick-feeding technique ³³P-labelled phosphoric acid was fed in situ to canola (Brassica napus) and lupin (Lupinus angustifolius) grown in sand or loam soils in sealed pots.

Results

Recovery of ³³P was 93 % in the plant-soil system and 7 % was sorbed to the wick. Significantly more ³³P was allocated below-ground than to shoots for both species with 59–90 % of ³³P measured in recovered roots plus bulk and rhizosphere soil. ³³P in recovered roots was higher in canola than lupin regardless of soil type. The proportion of ³³P detected in soil was greater for lupin than canola grown in sand and loam (37 and 73 % lupin, 20 and 23 % canola, respectively). Estimated total below-ground P accumulation by both species was at least twice that of recovered root P and was a greater proportion of total plant P for lupin than canola.

Conclusion

Labelling roots using ³³P via stem feeding can empower quantitative estimates of total below-ground plant P and root dry matter accumulation which can improve our understanding of P distribution in soil-plant systems.

     

No need to breed for enhanced colonization by arbuscular mycorrhizal fungi to improve low-P adaptation of West African sorghums

Background and aims

Herbaspirillum seropedicae (Hs) Z67 a diazotrophic endophyte was genetically engineered for secretion of 2-keto-D-gluconic acid by heterologous expression of genes for pqq synthesis and gluconate dehydrogenase to study its beneficial effect on plants.

Methods

Two plasmids, pJNK5, containing a 5.1 Kb pqq gene cluster of Acinetobacter calcoaceticus and pJNK6, carrying in addition the Pseudomonas putida KT2440 gluconate dehydrogenase (gad) operon were constructed in pUCPM18Gm^r under Plac promoter. H. seropedicae Z67 transformants were monitored for P and K solubilization, cadmium (Cd) tolerance and rice growth promotion.

Results

Hs (pJNK5) secreted 23.5 mM gluconic acid and Hs (pJNK6) secreted 3.79 mM gluconic acid and 15.8 mM 2-ketogluconic acid respectively. Under aerobic conditions, Hs (pJNK5) and Hs (pJNK6) solubilized 239.7 μM and 457.7 μM P on HEPES rock phosphate and, 76.7 μM and 222.7 μM K on HRPF (feldspar), respectively, in minimal medium containing 50 mM glucose. Under N free minimal medium, similar effects of P and K solubilization were obtained. Hs (pJNK5) and Hs (pJNK6) inoculation increased the biomass, N, P, K content of rice plants (Gujarat – 17). These plants also accumulated 0.73 ng/g PQQ, and had improved growth and tolerance to CdCl₂.

Conclusions

Incorporation of pqq and gad gene clusters in H. seropedicae Z67 imparted additional plant growth promoting traits of P and K solubilization and ability to alleviate Cd toxicity to the host plant.