Analog Implementation of the Softmax Function
Jacob Sillman

TL;DR
This paper introduces an analog circuit design for the Softmax function, demonstrating modular scalability and high accuracy through experimental and simulation results with BJT and NMOS technologies.
Contribution
It presents a novel modular analog implementation of Softmax that scales linearly and achieves high computational accuracy in both experimental and simulation settings.
Findings
Accuracy within 4.2% in prototype
Simulation accuracy within 1.3% (BJT)
Simulation accuracy within 0.9% (NMOS)
Abstract
An analog implementation of the Softmax activation function is presented. A modular design is proposed, scaling linearly with the number of inputs and outputs. The circuit behaves similarly using both a BJT and NMOS design scheme. Experimental results extracted from a BJT breadboard prototype presents computational accuracy within 4.2% margin of error. Simulation data presents accuracy within 1.3% margin of error for BJT and 0.9% margin of error for NMOS design schemes.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAnalog and Mixed-Signal Circuit Design · Advancements in Semiconductor Devices and Circuit Design · Neural Networks and Applications
MethodsSoftmax
