If we build a cognitive agent that works in a nonstationary dynamic environment, the agent needs to consider the ability to encode and update everyday experiences and recall or expect the exact data from memory. Moreover, lifelong experiences are composed of various types of contextual information and activities. Each attribute has a special relationship to other attributes. Sometimes,

Aurora Kinase pathway more than three attributes are synchronized to composite a situation. Therefore, we consider that the experience data need to be modeled in a temporal relationship between input events and even in a causal relationship between contextual values. From the previous theories on familiarity, a computational model for recognition memory is required to satisfy the human-like performance of old/new judgments. So far, however, the exact mechanism of human brain is not fully investigated. The relationship between familiarity and recollection is still controversial. Some important issues of recognition memory such as aging, forgetting, and context dependency are unsolved

yet. To approach the human-like cognitive model, the model may follow a neural mechanism like human or it may show a similar performance to human. In this paper, we try to show a similar performance of familiarity judgment while dealing with lifelong experience data to evaluate the model. If the suggested model reveals a human-like performance by comparing the ROC curves, we can further investigate the undisclosed characteristics of recognition memory from the memory model. To enable lifelong learning in recognition memory, we apply a

flexible structure to implement the computational memory model. Particularly for the role of familiarity judgment in recognition memory, we suggest a computational memory model that enables lifelong learning by applying a hypergraph structure. The model is built based on the concept of content-addressable memory. It records the data into the model without filtering or modifying. The hypergraph also supports a high-order relationship between nodes. By building a layered hypergraph structure, temporal events can be integrated into a network. Through the memory model, in this paper, we try to answer the following questions. Does the proposed memory model show a human-like Entinostat ROC performance of familiarity judgment? What is the remarkable characteristic of the proposed recognition memory model for treating lifelong experience data? Does the memory model maintain the performance of both familiarity judgment and pattern completion under nonstationary encoding conditions? In the following sections, we introduce the hypergraph-based memory mechanism and evaluate the memory model, which shows a similar performance to the human tasks of recognition judgment.