ABSTRACT
In addition to the short-term dynamics that occur during retrieval, the SCALIR network also exhibits long-term changes — in the form of weight modification and link creation — that affect its behavior over many queries and many users. This learning occurs at run time, in response to users’ browsing behavior. This chapter includes a range of discussions about learning issues, from a general examination of learning strategies for information retrieval to the tuning of specific parameters in candidate learning rules. Although they differ in scope and detail, they can all be viewed as successively refined versions of the question “how can SCALIR learn?”
