View source for 18165802. CONFIGURABLE NONLINEAR ACTIVATION FUNCTION CIRCUITS simplified abstract (QUALCOMM Incorporated)

=CONFIGURABLE NONLINEAR ACTIVATION FUNCTION CIRCUITS=


==Organization Name==

[[:Category:QUALCOMM Incorporated|QUALCOMM Incorporated]]


[[Category:QUALCOMM Incorporated]]

==Inventor(s)==

[[:Category:Ren Li of San Diego CA (US)|Ren Li of San Diego CA (US)]][[Category:Ren Li of San Diego CA (US)]]

[[:Category:Prajakt Kulkarni of San Diego CA (US)|Prajakt Kulkarni of San Diego CA (US)]][[Category:Prajakt Kulkarni of San Diego CA (US)]]

[[:Category:Suren Mohan of San Diego CA (US)|Suren Mohan of San Diego CA (US)]][[Category:Suren Mohan of San Diego CA (US)]]

[[:Category:Aaron Douglass Lamb of San Diego CA (US)|Aaron Douglass Lamb of San Diego CA (US)]][[Category:Aaron Douglass Lamb of San Diego CA (US)]]

==CONFIGURABLE NONLINEAR ACTIVATION FUNCTION CIRCUITS - A simplified explanation of the abstract==


This abstract first appeared for US patent application 18165802 titled 'CONFIGURABLE NONLINEAR ACTIVATION FUNCTION CIRCUITS

==Simplified Explanation==

The patent application describes a method for processing input data using configurable nonlinear activation function circuits. Here is a simplified explanation of the abstract:

* The method involves using configurable nonlinear activation function circuits to process input data.
* One or more first configurable nonlinear activation function circuits are used to generate an exponent output by performing an exponential function on the input data.
* The exponent outputs from the first circuits are summed together.
* A second configurable nonlinear activation function circuit is used to process the summed exponent output and generate an approximated log softmax output by performing a natural logarithm function.

Potential Applications:

* This technology can be applied in various fields where processing input data using nonlinear activation functions is required.
* It can be used in machine learning algorithms, deep learning models, and neural networks for tasks such as image recognition, natural language processing, and data analysis.

Problems Solved:

* The method provides a way to efficiently process input data using configurable nonlinear activation function circuits.
* It allows for the approximation of log softmax outputs, which can be useful in various mathematical and statistical calculations.

Benefits:

* The method offers flexibility by using configurable nonlinear activation function circuits that can be adjusted or programmed for different functions.
* It provides a more efficient way to perform exponential and natural logarithm functions on input data.
* The approximated log softmax outputs can be used in various mathematical calculations and statistical analysis.


==Original Abstract Submitted==

Certain aspects of the present disclosure provide a method for processing input data by a set of configurable nonlinear activation function circuits, including generating an exponent output by processing input data using one or more first configurable nonlinear activation function circuits configured to perform an exponential function, summing the exponent output of the one or more first configurable nonlinear activation function circuits, and generating an approximated log softmax output by processing the summed exponent output using a second configurable nonlinear activation function circuit configured to perform a natural logarithm function.



[[Category:G06F7/556]]
[[Category:G06F7/50]]