Fast IDS Computing System Method and its Memristor Crossbar-based Hardware Implementation
Active Learning Method (ALM) is one of the powerful tools in soft computing that is inspired by human brain capabilities in processing complicated information. ALM, which is in essence an adaptive fuzzy learning method, models a Multi-Input Single-Output (MISO) system with several Single-Input Single-Output (SISO) subsystems. Ink Drop Spread (IDS) operator, which is the main processing engine of this method, extracts useful features from the data without complicated computations and provides stability and convergence as well. Despite great performance of ALM in applications such as classification, clustering, and modelling, an efficient hardware implementation has remained a challenging problem. Large amount of memory required to store the information of IDS planes as well as the high computational cost of the IDS computing system are two main barriers to ALM becoming more popular. In this paper, a novel learning method is proposed based on the idea of IDS, but with a novel approach that eliminates the computational cost of IDS operator. Unlike traditional approaches, our proposed method finds functions to describe the IDS plane that eliminates the need for large amount of memory to a great extent. Narrow Path and Spread, which are two main features used in the inference engine of ALM, are then extracted from IDS planes with minimum amount of memory usage and power consumption. Our proposed algorithm is fully compatible with memristor-crossbar implementation that leads to a significant decrease in the number of required memristors (from O(n^2) to O(3n)). Simpler algorithm and higher speed make our algorithm suitable for applications where real-time process, low-cost and small implementation are paramount. Applications in clustering and function approximation are provided, which reveals the effective performance of our proposed algorithm.