This thesis mainly introduces two topics in complexity science: Ranking of
researchers and optimal learning in complex networks, both of which help us
understand the ever-increasing complexity of the world. First, we design a novel
method for ranking or evaluating researchers in citation networks, which is very
vital for the scientific community. A well-designed ranking method can be used
to rank scientists in various practical tasks, such as hiring, funding application
and promotion. However, a large number of evaluation methods are designed
based on citation counts which can merely evaluate scientists’ scientific
impact but can not evaluate their innovation ability which actually is a crucial
characteristic for scientists. In addition, when evaluating scientists, it has
become increasingly common to only focus on their representative works
rather than all their papers. Accordingly, we here propose a hybrid method by
combining scientific impact with innovation under representative works
framework to rank scientists. Our results show that the ranking performance of
the hybrid method is the best compared with other mainstream methods. This
study can provide policy makers an effective way to rank scientists from more
comprehensive dimensions.
Besides, taking into account the time bias issue in ranking or evaluating the
long-term or lifetime achievements of researchers, a competition-aware
ranking method for researchers is proposed. Since the number of scholars and
the number of scholarly outputs grow exponentially with time, a well-designed
unbiased ranking metric for researchers is of prime importance. To rank
scholars, it is important to put their achievements in perspective by comparing
them with the achievements of other scholars active in the same period. We
propose here a particular way of doing so: by computing the evaluated
scholar's share on each year's citations which quantifies how the scholar fares
in competition with the others. Our results show that the new ranking method
significantly outperforms other ranking methods in identifying the prize
laureates.
Second, we study the optimal learning in information networks. Similar to
ranking for researchers, optimal learning is also a significant topic in complexity
science. The ever-increasing complexity of the world around us challenges our
ability to understand it. We study a model where a single agent learns node
types or say forms opinions about many interconnected topics. This model was
shown to be challenging for a boundedly rational agent. We extend it by
assuming that the agent combines a targeted study of some topics and various
heuristics for the remaining ones. We find that the highest opinion accuracy is
generally achieved neither when one topic is studied very well nor when many
topics are studied a little. Despite big differences in accuracy between the
considered heuristics, the optimal number of topics in which the study budget
is invested grows linearly with the budget for all of them. The study budget
necessary to achieve the desired opinion accuracy exhibits a simple scaling
with the total number of topics. In this way, we exemplify how to use limited
cognitive effort for efficient learning in a complex system.