Learning Optimal Decision Trees Using Caching Branch-and-Bound Search

Authors: Gaël Aglin, Siegfried Nijssen, Pierre Schaus3146-3153

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In our experiments we focus our attention on traditional decision tree learning problems with little other constraints... Table 2 shows the results for a maximum depth equal to 4... The difference in performance is further illustrated in Figure 4, which gives cactus plots for each algorithm, for different depth constraints... To answer Q4, we repeat these tests on continuous data.
Researcher Affiliation Academia Ga el Aglin, Siegfried Nijssen, Pierre Schaus firstname.lastname@uclouvain.be ICTEAM, UCLouvain Louvain-la-Neuve, Belgium
Pseudocode Yes Algorithm 1: DL8(maxdepth, minsup); Algorithm 2: DL8.5(maxdepth, minsup)
Open Source Code Yes Our implementation is publicly available at https://github. com/aglingael/dl8.5 and can easily be used from Python.
Open Datasets Yes We compare our algorithms on 24 binary datasets from CP4IM5, described in the first columns of Table 2. ... These datasets were obtained from the UCI repository6 and are summarized in the first columns of Table 3. ... 5https://dtai.cs.kuleuven.be/CP4IM/datasets/, 6https://archive.ics.uci.edu/ml/index.php
Dataset Splits No We do not split our datasets in training and test sets since the focus of this work is on comparing the computational performance of algorithms that should generate decision trees of the same quality.
Hardware Specification Yes Experiments were performed on a server with an Intel Xeon E5-2640 CPU, 128GB of memory, running Red Hat 4.8.5-16.
Software Dependencies Yes The implementations of Bin OCT1, DL8 and the CP-based approach2 used in our comparison were obtained from their original authors, and use the CPLEX 12.93 and Osca R 4 solvers.
Experiment Setup Yes We run the different algorithms for 10 minutes on each dataset and for a maximum depth of 2, 3 and 4. All the tests are run with a minimum support of 1 since this is the setting used in Bin OCT.