Fixed-Point Exponential Calculation Circuit for Low-Power SoftMax
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating the softMax function in deep learning require high power consumption due to extensive exponential computations, which are necessary for maintaining calculation accuracy.
Innovation Solution
The method involves converting input values to a fixed point number format, using table lookup and interpolation to calculate exponential functions, thereby reducing the need for multipliers and adders, and includes a circuit design with lookup and interpolation circuits to perform these calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional exponential computation methods are used to maintain calculation accuracy in softMax function, then calculation precision is improved, but power consumption increases significantly
Solution Approach 1:
The input value is segmented into integer and decimal parts, allowing the exponential computation to be divided into table lookup for integer part and interpolation for decimal part, reducing the need for complex multipliers and adders
Solution Approach 2:
Pre-computed exponential values are stored in lookup tables, replacing real-time exponential computation with table lookup operations that consume less power while maintaining accuracy through interpolation
2Measurement precision
If more multipliers and adders are used to achieve higher calculation accuracy for exponential functions, then measurement precision is improved, but device complexity increases
Solution Approach 1:
Lookup tables store pre-computed exponential values, replacing complex multiplier and adder circuits with simpler memory access and interpolation logic, reducing device complexity while maintaining precision
Solution Approach 2:
By separating integer and decimal processing, the circuit uses simple lookup for integer part and lightweight interpolation for decimal part, avoiding the need for complex full-precision exponential computation circuits
Data Source
AI summary
Disclosed are exponential function calculation method, exponential function calculation system, and exponential function calculation circuit for calculating an exponential function value of an input value. The exponential function calculation method includes: dividing the input value as a sum of an integer value and a decimal value; representing the integer value by an integer bit string with a first length, and representing the decimal value by a decimal bit string with a second length; looking up a first table to obtain a first value corresponding to an exponent function value of the integer value according to the integer bit string, and interpolating to obtain an interpolation value corresponding to an exponent function value of the decimal value according to the decimal bit string; and calculating an output value corresponding to an exponent function value of the input value according to a product of the first value and the interpolation value.