We propose a new classification for multi-agent learning algorithms, with each league of players characterized by both their possible strategies and possible beliefs. Using this classification, we review the optimality of existing algorithms and discuss some insights that can be gained. We propose an incremental improvement to the existing algorithms that seems to achieve average payoffs that are at least the Nash equilibrium payoffs in the long-run against fair opponents. / Singapore-MIT Alliance (SMA)

A fundamental goal in ecology and evolutionary biology has been to understand how microevolutionary forces affect the origin and maintenance of mutualisms over ecological and evolutionary time scales. Mutualistic partners vary in the reciprocal benefits they provide, yet the role of partner variation on microevolutionary forces that impact the maintenance of mutualisms is unclear. Using the mutualism between legumes and nitrogen fixing symbionts, my dissertation investigated the ecological and evolutionary consequences of variation in partner quality. In the first experiment, I demonstrate how insect herbivory can change the costs and benefits of associating with exploiters, and that some degree of exploitation from non-beneficial rhizobia can reduce insect herbivory, thus removing the fitness advantage of associating purely with beneficial rhizobia. In the second study, I examine how rhizobia genotype modifies competition between hosts grown in kin and non-kin groups. I show that lower fitness in plant kin groups can simply be a by-product of genetic variance in plant size and non-linear relationships between plant size and fitness. I further show that the symbiotic community can change difference in fitness between kin and non-kin groups independent of these by-product effects. In my last chapter, I provide the first empirical evidence that an important mechanism for mutualism stability-- the ability for hosts to preferentially associate with beneficial rhizobia-- is genetically variable and can evolve in response to exploitation. I also show that host preference for beneficial rhizobia can be maintained in legume populations, even in the absence of exploitation. My dissertation provides insight into the potential evolutionary dynamics of stabilizing mechanisms by suggesting that the agents of selection that affect the level of host exploitation can come from biotic factors other than the exploiters themselves. My dissertation has also shown that inclusion of other ecological interactions, such as herbivory, can provide valuable perspective on fitness effects of symbionts on their hosts, and can even change our fundamental assumptions about the effects of exploitation on host fitness, which has formed the backbone of mutualism theory.

A fundamental goal in ecology and evolutionary biology has been to understand how microevolutionary forces affect the origin and maintenance of mutualisms over ecological and evolutionary time scales. Mutualistic partners vary in the reciprocal benefits they provide, yet the role of partner variation on microevolutionary forces that impact the maintenance of mutualisms is unclear. Using the mutualism between legumes and nitrogen fixing symbionts, my dissertation investigated the ecological and evolutionary consequences of variation in partner quality. In the first experiment, I demonstrate how insect herbivory can change the costs and benefits of associating with exploiters, and that some degree of exploitation from non-beneficial rhizobia can reduce insect herbivory, thus removing the fitness advantage of associating purely with beneficial rhizobia. In the second study, I examine how rhizobia genotype modifies competition between hosts grown in kin and non-kin groups. I show that lower fitness in plant kin groups can simply be a by-product of genetic variance in plant size and non-linear relationships between plant size and fitness. I further show that the symbiotic community can change difference in fitness between kin and non-kin groups independent of these by-product effects. In my last chapter, I provide the first empirical evidence that an important mechanism for mutualism stability-- the ability for hosts to preferentially associate with beneficial rhizobia-- is genetically variable and can evolve in response to exploitation. I also show that host preference for beneficial rhizobia can be maintained in legume populations, even in the absence of exploitation. My dissertation provides insight into the potential evolutionary dynamics of stabilizing mechanisms by suggesting that the agents of selection that affect the level of host exploitation can come from biotic factors other than the exploiters themselves. My dissertation has also shown that inclusion of other ecological interactions, such as herbivory, can provide valuable perspective on fitness effects of symbionts on their hosts, and can even change our fundamental assumptions about the effects of exploitation on host fitness, which has formed the backbone of mutualism theory.

Memoria para optar al título de Geógrafo / En los últimos años la ciudad de Pichilemu ha experimentado un sostenido crecimiento físico y demográfico por influencia de la actividad turística, ocasionando una presión creciente sobre áreas de alto valor ecológico, y por ende, alteraciones en la biodiversidad local. Siguiendo la tendencia de América Latina, Chile posee un déficit de estudios de ecología urbana, motivo por el cual esta memoria contribuye evaluando la biodiversidad de avifauna y vegetación leñosa en las coberturas de suelo del mosaico urbano de Pichilemu. El levantamiento de información de biodiversidad se llevó a cabo mediante conteo de aves e inventarios florísticos entre el período estival del 2017 y el de 2018. Con esta información se calculó riqueza, abundancia, tres índices de diversidad alfa (Shannon-Wiener, Pielou y Simpson), en conjunto con los porcentajes de especies endémicas, nativas y amenazadas. Además, se analizó si la distancia entre los puntos de muestreo y los componentes de infraestructura verde diversos en Shannon influencia la distribución espacial de los valores de los parámetros de diversidad alfa. Paralelamente se clasificaron las especies de avifauna en gremios tróficos y según su grado de sensibilidad a la urbanización. Asimismo, se compararon las coberturas de suelo en función de su biodiversidad a través del test de Kruskall-Wallis y el test de Dunn. Por último, se realizó un análisis de clúster para ver la similitud entre las coberturas de suelo. Los resultados señalan un patrón bimodal en el índice de Shannon, pues existen dos grupos de coberturas de suelo que presentan los valores máximos en esta variable en Pichilemu. El primero de ellos, con los valores más altos del índice de Shannon, está compuesto por coberturas naturales tales como los humedales y cuerpos de agua, quebradas y cursos de agua, y matorral arborescente. El otro grupo, con los segundos valores más altos del índice de Shannon, está integrado por coberturas híbridas como el tejido urbano continuo y el tejido urbano discontinuo. En Pichilemu, las especies de avifauna son mayoritariamente nativas, mientras que gran parte de las especies de vegetación leñosa son exóticas, muchas de ellas siendo exóticas invasoras. A pesar de su alta diversidad, el tejido urbano continuo alberga casi todos los individuos pertenecientes a una especie urban exploiter (Passer domesticus). Al contrario, los individuos de las especies urban avoiders se concentran en coberturas diversas y poco intervenidas como los humedales y cuerpos de agua, quebradas y cursos de agua, playas, dunas y arenales, y el matorral arborescente. Al mismo tiempo, estas coberturas naturales son las que exhiben la mayor cantidad de gremios tróficos. Respecto al análisis de correlación, la distancia entre los puntos de muestreo y las coberturas correspondientes a componentes de infraestructura verde diversos en Shannon no explica totalmente la variabilidad de los parámetros de diversidad alfa. Mediante el análisis de clúster, se destacan los altos niveles de diversidad del conglomerado compuesto por los humedales y cuerpos de agua, quebradas y cursos de agua, matorral arborescente, matorral pradera, y las plantaciones, De este modo, se concluye que estas coberturas deberían ser priorizadas para las labores de restauración ecológica y conservación de la biodiversidad. / Throughout the last years, the city of Pichilemu has undergone a sharp spatial and demographic growth influenced partly by tourist activity, which it caused an increasing pressure on high ecological value areas and local biodiversity. Following the same trend as in Latin America, Chile has few studies on urban ecology; hence this study contributes with an assessment of avifauna and woody vegetation biodiversity in land covers belonging to the urban mosaic of Pichilemu. All field samplings were conducted during 2017 and 2018 summer times using bird counting and floristic inventories. Richness, abundance, alpha diversity indexes (Shannon-Wiener, Pielou and Simpson) of species were calculated, along with the proportion of endemic, native and threatened species. Moreover, the distance between sampling points and diverse components of the green infrastructure of Pichilemu city was analysed to demonstrate whether it influences the spatial distribution of alpha diversity parameters values. Avifauna species were classified into trophic guilds and three distinct categories (urban exploiter, urban adapter and urban avoider) basing on their response to urbanisation. Furthermore, in order to compare land covers regarding their biodiversity, Kruskal-Wallis and Dunn's test were applied. Finally, a cluster analysis was carried out to see the similarity among land covers and sampling points. The results show a bimodal distribution pattern on the Shannon index, since there are two groups of land covers with maximum values both in avifauna and woody vegetation in Pichilemu. The most diverse group in Shannon index is made up of natural land covers such as wetlands and water bodies, streams and water courses and tree-shaped scrubland. The second most diverse group in Shannon index is made up of hybrid land covers such as continuous urban fabric and discontinuous urban fabric. Most of avifauna species are native whilst the majority of woody vegetation species are exotic in Pichilemu. Even though continuous urban fabric is diverse, almost all individuals of avifauna belong to an urban exploiter species (Passer domesticus). On the other hand, individuals belonging to urban avoider species are prone to be found in diverse and less disturbed land covers such as wetlands and water bodies, streams and water courses, strands, dunes and sandbanks and tree-shaped scrubland. Likewise, these land covers contain the highest number of trophic guilds of avifauna. Regarding correlation analysis, the distance between sampling points and diverse green infrastructure components does not explain completely the spatial variability of alpha diversity parameters. Through cluster analysis, cluster comprised of wetlands and water bodies, streams and water courses, tree-shaped scrubland, shrub brush and plantations has outstanding diversity values. Therefore, these land covers should be prioritised for ecological restoration and biodiversity conservation measures.

No description available.