Memcomputing: a computing paradigm to store and process information on the same physical platform (1211.4487v2)

Abstract: In present day technology, storing and processing of information occur on physically distinct regions of space. Not only does this result in space limitations; it also translates into unwanted delays in retrieving and processing of relevant information. There is, however, a class of two-terminal passive circuit elements with memory, memristive, memcapacitive and meminductive systems -- collectively called memelements -- that perform both information processing and storing of the initial, intermediate and final computational data on the same physical platform. Importantly, the states of these memelements adjust to input signals and provide analog capabilities unavailable in standard circuit elements, resulting in adaptive circuitry, and providing analog massively-parallel computation. All these features are tantalizingly similar to those encountered in the biological realm, thus offering new opportunities for biologically-inspired computation. Of particular importance is the fact that these memelements emerge naturally in nanoscale systems, and are therefore a consequence and a natural by-product of the continued miniaturization of electronic devices. We will discuss the various possibilities offered by memcomputing, discuss the criteria that need to be satisfied to realize this paradigm, and provide an example showing the solution of the shortest-path problem and demonstrate the healing property of the solution path.

• The paper proposes memcomputing as a novel paradigm overcoming classical computing limitations, like the von Neumann bottleneck, by integrating storage and processing on the same platform with memelements.
• Demonstrating the paradigm, the paper shows a memristive network that solves the shortest-path problem, illustrating collective dynamics, dynamic adaptation, and self-healing capabilities.
• Memcomputing offers practical implications such as high integration density and energy-efficient nanoscale processing, poised for immediate implementation and enabling future neuromorphic computing advancements.

Memcomputing: Unifying Storage and Computation on the Same Physical Platform

Memcomputing, as proposed by Di Ventra and Pershin, represents a paradigm shift in information processing by integrating storage and computing on a single physical platform through the use of memelements, such as memristors, memcapacitors, and meminductors. These two-terminal components possess memory capabilities and analog computing capacities inherently suited to nanoscale systems. The paper discusses the memcomputing concept, criteria for implementation, and potential applications, particularly focusing on the shortest-path problem.

The authors of the paper address the fundamental limitations of classical computing architectures, especially the von Neumann bottleneck, which segregates storage and processing into distinct physical regions, leading to inefficient data transfer and latency. The integration of storage and computation provided by memelements can potentially alleviate these limitations. By mimicking the biological adaptability seen in human neural networks, memelements offer an avenue for biologically-inspired computing, segueing into a new frontier of technology development.

Memcomputing Criteria

The paper lays down six criteria crucial for effective memcomputing implementation:

1. Scalability and Integration: Emphasized is the need for scalable, massively-parallel architectures that marry information storage and processing.
2. Information Storage Durability: Memelements should retain information substantially beyond the duration of computational processes, ideally manifesting non-volatility.
3. Initialize Memory States: Effective mechanisms should be in place for the initialization of memelements to prepare them for computations.
4. Collective Dynamics and Strong Memory Content: The ability of memelements to interact collectively is imperative, ensuring robust computation through significant differences in memory states.
5. Final Result Readout: Techniques for non-invasive memory state readout post-computation are crucial for uninterrupted processing.
6. Imperfection and Noise Insensitivity: Architectures should demonstrate resilience against minor physical imperfections and fluctuations.

Results and Implications

A key illustration of memcomputing is provided through the solution of the shortest-path problem using a memristive network. The system dynamically evolves to adjust its memristance based on past activities, learning and adapting in a manner similar to biological systems such as ant colonies. This system sees the full utilization of all six criteria in practical application, notably exhibiting self-healing in computational paths subject to damage, showcasing its robust, adaptive nature.

The numerical simulations underscore the system's gradual optimization as memristive states interact collectively to form the shortest path. The concept of network entropy, as introduced in the paper, further delineates the efficiency and evolution of these networks as computations proceed—a significant decrease in entropy indicates fewer available paths and optimized computation.

Practical and Theoretical Implications

Memcomputing presents compelling theoretical implications for computing architectures, potentially leading to systems analogous to cognition, with self-adaptive and healing capabilities. Practically, it promises heightened integration densities and energy-efficient processing at nanoscale dimensions, underpinning the future development of compact, high-performance computing systems.

Conclusion

This research asserts memcomputing as a viable, potentially transformative approach to overcoming current computational inefficiencies. With memelements naturally emerging at the nanoscale, this paradigm offers immediate implementation possibilities, circumventing the prolonged development associated with other computing strategies, such as quantum computing. Memcomputing stands poised as a key enabler for advancements in neuromorphic and biologically-inspired computing architectures. As a nascent area of exploration, further research into the design and optimization of memcomputing architectures continues to hold promise for future developments in computing intelligence.