Memcomputing

Memcomputing is a new computing paradigm, inspired by the operation of our own brain, where the whole burden of computation is directly put into the memory. Like the brain, memcomputing machines compute with and in memory without the need of a separate CPU. The memory allows learning and adaptive capabilities, bypassing broken connections and self-organizing the computation into the solution path, much like the brain is able to sustain a certain amount of damage and still operate seamlessly. The whole concept of memcomputing can be realized in practice utilizing physical properties of many materials and systems which show a certain degree of time nonlocality (memory) in their response functions at particular frequencies and strengths of the inputs. There are 3 main features unique to memcomputing machines: intrinsic parallelism, i.e. all memelement processors operate simultaneously and collectively during the computation, functional polymorphism, i.e. the ability of computing different functions without modifying the topology of the machine network by simply applying the appropriate input signals, and information overhead, i.e. the capability of an interacting memprocessor network to store and compress an information amount larger than that possible by the same noninteracting memprocessors.

ERC Sector:

  • PE6_4 Theoretical computer science, formal methods, and quantum computing

Keywords:

  • Neural computing
  • In-memory computing

Gruppi di ricerca

Persone di riferimento