A current-mode CMOS loser-take-all with minimum function for neural computations

A Current-Mode Analog Circuit for Reinforcement Learning Problems

“…An analog voltage-mode implementation using the binary relation inference network (BRIN) [10] for solving the dynamic programming was previously reported. In line with the previous work, with a simpler and more effective currentmode design using only loser-take-all (LTA or min) circuits [5], [14] previously developed for neural-fuzzy systems, to solve the shortest path problem.…”

Low Voltage Lazzaro's WTA with enhanced loop gain

“…Previous efforts to reduce this high voltage requirement have been made. In [2] using SOI technology, an extra complementary source follower stage is introduced in each cell loop; this fact allows the circuit to work as a Loser Take All (LTA) with a reduction on the voltage requirement. Nevertheless, the limited gain ∼0.8 of source followers reduces the loop gain of the cell in this approach, so, there is just a marginal improvement on accuracy over the original Lazzaro's circuit.…”

A novel low voltage CMOS controllable fuzzy inference circuit

“…In some cases such as FLC, to increase the controllability of circuit, two operations are required. Using circuits of WTA circuit to obtain Max operation [6,11,18,21] and subtract from fixed value to obtain LTA and Min operation [4,6,18,19] loose the accuracy and DC range of input [5]. The analog subtraction causes loss of accuracy and limits the input current to the value of a fixed reference, moreover, it needs large number of MOS transistors, which increases the area and power consumption [24].…”