BINGO: brain-inspired learning memory

Abstract Storage and retrieval of data in a computer memory play major role system performance. Traditionally, organization is ‘static’—i.e. it does not change based on the application-specific characteristics access behaviour during operation. Specifically, case content-operated (COM), association block with search pattern (or cues) granularity (details) stored do evolve. Such static nature memory, we observe, only limits amount can store given physical storage, but also misses opportunity for performance improvement various applications. On contrary, human characterized by seemingly infinite plasticity storing retrieving data—as well as dynamically creating/updating associations between corresponding cues. In this paper, introduce BINGO, brain-inspired learning paradigm that organizes flexible neural network. network structure, strength associations, adjust continuously operation, providing unprecedented benefits. We present associated storage/retrieval/retention algorithms which integrate formalized process. Using an operational model, demonstrate BINGO achieves order magnitude times effective storage capacity using CIFAR-10 dataset wildlife surveillance when compared to traditional memory.