3 Techniques of Tree Sampling

At the 2015 University Day organized in the Systems workshop organized by Wang Chen at IBM Toronto, José Nelson Amaral of University Alberta gave a talk explaining the following studies.  The problem is estimating the size of a tree by traversing some but not all nodes.

“Mathematics of computation” by Donald Knuth in American Mathematical Society 1975 gave a sampling solution: Go down random tree paths to the leaves, and take the average of the sampled sub-tree size as the actual average sub-tree size.

Heuristic sampling by Peng Chen in 1992 takes samples but reuses the past results.  It uses the term Stratification but the technique sounds like memoization.  An example is the Fibonacci series.

The Alberta work is to estimate the number of leaf tasks in a recursive fork-join (Cilk) program.  The solution is a complete exploration of the top of the tree until it has the nodes 10 times the number of processors.  Then it uses the Knuth method to estimate the size of a sub-tree.

Another example is 3×3 puzzle, 181,440 states and 4×4, more than 10^13 states.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s