|
|
|
Quick Overview
|
|
GPU's higher processing power compared to a standard CPU comes
at the cost of reduced data caching and flow control logic as
more transistors have to be devoted to data processing. This
imposes certain limitations in terms of how an application
may access memory and implement flow control. As a result,
implementation of certain algorithms (even trivial ones) on the
GPU may be difficult or may not be computationally justified.
|
|
Histogram has been traditionally difficult to compute efficiently
on the GPU. Lack of an efficient histogram method
on the GPU, often requires the programmer to move the data
back from the device (GPU) memory to the host (CPU), resulting
in costly data transfers and reduced efficiency. A simple
histogram computation can indeed become the bottleneck of
an otherwise efficient application.
|
Download the Code
|
You can find the source code for two efficient histogram computation
methods for CUDA compatible GPUs here.
The methods are described in the following publications:
"Efficient histogram algorithms for NVIDIA CUDA compatible devices"
and "Speeding up mutual information computation using NVIDIA CUDA hardware".
|
|
|
|
|
Quick Overview
|
|
Mutual information of two random variables is the amount of
information that each carries about the other and is defined as
where H(X|Y) is the information content of random variable X if
Y is known, H(X,Y) is the joint entropy of the two random
variables and is a measure of combined information of the two random
variables. I(X;Y) can be thought of as the reduction in
uncertainty of random variable X as a result of knowing Y. The
uncertainty is maximally reduced, when there is a one-to-one mapping
between the two random variables and is not reduced at all if the
two random variables are independent and do not provide any
information about one another.
The concept of mutual information has its origins in information theory and
is widely used in other disciplines. In medical image analysis, mutual information
is used as a similarity measure for multi-modal registration.
|
Download the Code
|
You can find the source code for efficient MI computation
methods for CUDA compatible GPUs here.
The methods are described in the following publication:
"Speeding up mutual information computation using NVIDIA CUDA hardware".
|
|
|