Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2016-12-15T18:17:07Z No. of bitstreams: 1 RebeccaDeMouraDinizGomes_DISSERT.pdf: 2411622 bytes, checksum: 6b5e6ef2c6fd430fe0ff200b6352cd44 (MD5) / Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2016-12-20T21:33:44Z (GMT) No. of bitstreams: 1 RebeccaDeMouraDinizGomes_DISSERT.pdf: 2411622 bytes, checksum: 6b5e6ef2c6fd430fe0ff200b6352cd44 (MD5) / Made available in DSpace on 2016-12-20T21:33:44Z (GMT). No. of bitstreams: 1 RebeccaDeMouraDinizGomes_DISSERT.pdf: 2411622 bytes, checksum: 6b5e6ef2c6fd430fe0ff200b6352cd44 (MD5) Previous issue date: 2016-05-27 / Grande tem sido o interesse nas difus?es an?malas, pois se apresentam nas mais diversas ?reas do conhecimento. A introdu??o de perfil de mem?ria no caminhante aleat?rio torna-o numa din?mica estoc?stica n?o-markoviana, cujas correla??es criam superdifus?o, persistencia e log-periodicidade. Apresentamos uma revis?o da literatura sobre os perfis de mem?ria e introduzimos nosso modelo. O modelo de mem?ria binomial pode selecionar diferentes regi?es de perda de mem?ria, desde a inicial at? a recente. Dessa forma, investigamos o impacto da posi??o da perda de mem?ria no comportamento superdifusivo do caminhante aleat?rio e unificamos muitos dos resultados da literatura. Obtivemos que mem?rias iniciais geram maior superdifus?o medidas pelo coeficiente de Hurst, enquanto que mem?rias recentes tendem a diminuir a superdifus?o, tornando mais caminhantes adeptos da difus?o normal. Tamb?m investigamos o regime de mem?ria curta inicial, com largura tendendo a zero. Observamos log-periodicidade para alguns caminhantes sugerindo regimes diferentes de comportamento log-periodico, incluindo aqueles considerados de difus?o normal. Uma particularidade do modelo binomial s?o os resutados extremamente sim?tricos para o diagrama Hxr. / Great has been the interest in anomalous diffusion because they are present in several areas of knowledge. The introduction of a memory profile in random walk environment give them a non-Markovian stochastic dynamics, whose temporal correlations may create superdiffusion, persistence and log-periodicity. We present an overview of memory profile literature and introduce our model. The binomial memory model can select different memory loss regions, from the old to the recent one. Thus, we investigate the impact of memory loss location on superdiffusive behavior of a random walker and unify some literature results. We verify that old memory generates higher superdiffusion measured by the Hurst coefficient, while recent memory tends to decrease superdiffusion, causing more walkers to undergo normal diffusion. We also investigate the short initial memory region, with zero tending standard deviation. We observe log-periodicity for some walkers suggesting different regions of log-periodic behavior, including those considered as normal diffusion. A particularity of the binomial model is an extremely symmetric result to Hxr diagram.