Bidirectional associative memories

Abstract

Bidirectional Associative Memories (BAM) are systems that allow to associate pairs of patterns. Once a memory has learned, patterns can be recalled in two directions. BAMs have many applications in pattern recognition and image processing. The aim of this survey is to present several models of BAM throughout time, since Kosko [1988] proposed the first model; followed by those works based on or inspired by it, trying to improve recall capacity; to some recent one-shot models—such as Morphological BAM and Alpha-Beta BAM—which are of particular interest, given their superior performance.

Activity centered Design of Smart Phone User Interface: Learning App Execution Patterns with Neural Network Model

“…This might be viewed as an activity centered user interface design and such philosophy can be extended to other applications of smart phone user interface design. A simple feasibility test is done with the original BAM [9] but, for the scalability of the research, one may use other extended forms of BAM with respect to different applications [10].…”

Capacity of bidirectional associative memory

“…[1,2]). One of the key issues for BAM researchers is capacity [3]. This issue was examined a lot, but not completely (see e.g.…”

“…In our paper we continue to examine capacity of BAM. It should be noted that there are different modifications of BAM [3]. We use the original variant from [7].…”

Productivity Comparison among Software Projects

“…In accordance with the productivity value, a software project could be classified in very low, low, medium, or high; whereas regarding prediction models, statistical regression and machine learning models can be proposed to predict SP. Classifiers will be based on Bidirectional Associative Memories [18], ontologies [19], Alpha-Beta models [20], and associative models [21], whereas models such as [22] [23] [24] and [24] will be applied for SP prediction.…”