benchmark/valgrind/categorical_binary_vs_linear/Readme.md

In schalkdaniel/compboost: Efficient Component-Wise Boosting Implementation

Profiling Categorical Binary Base-Learner

Data Simulation

• 5x10e5 observations
• 20 classes of equal size

Therefore, we have a size of 76.29 for the raw data in binary form.

Considerations Binary vs. Linear Base-Learner

• Binary stores data as sparse matrix, linear base-learner does not

Valgrind

Profiling is done with massif for stack and heap. To visualize results we use the massif-visualizer.

Note that running R in debugging mode with massif speeds down the execution 20 to 50 times.

schalkdaniel/compboost documentation built on April 15, 2023, 9:03 p.m.