Profiling the fault tolerance for the adaptive protein processing associative memory

The Protein Processor Associative Memory (PPAM) is a novel hardware architecture for a distributed, decentralised bidirectional, hetero-associative memory, that can adapt online to changes in the training data. The PPAM is fundamentally different from traditional processing methods that tend to use arithmetic operations to perform computation. In this paper, we evaluate the fault tolerant properties of the PPAM and show that, given a sufficiently large number of Conflict-Resolving nodes in the network, faults result in graceful degradation and individual nodes are not critical to the correct operation of the whole.