Nov 5 at 12pm (GHC 6115) -- Dravy Sharma (TTIC) -- Learning accurate and interpretable decision trees

[Victor Akinwande]

Dear all,

We look forward to seeing you next *Tuesday (11/05) from 12:00-1:00 PM (ET)*
for the next talk of this semester's CMU AI Seminar, sponsored by SambaNova
Systems <https://sambanova.ai>. The seminar will be held in *GHC 6115 *with
pizza provided and will be streamed on Zoom.

To learn more about the seminar series or to see the future schedule,
please visit the seminar website: http://www.cs.cmu.edu/~aiseminar/.

Next Tuesday (11/05), Dravy Sharma (TTIC) will be giving a talk titled
"Learning accurate and interpretable decision trees".

*Abstract*: Decision trees have remained popular in machine learning and
data science over the past several decades for being effective, "white-box"
models, and have often been treated as a gold standard for
interpretability. While learning even approximately optimal decision trees
is known to be computationally hard in the worst case, we present a
positive perspective for selecting the learning algorithms that yield the
most accurate and interpretable decision trees for repeated data instances
coming from the same domain.

We propose a novel family of splitting criteria used in top-down decision
tree learning, which unifies entropy, Gini impurity and the theoretically
motivated Kearns-Mansour criterion, and show how to learn the best best
splitting function for the data at hand. We also study the sample
complexity of tuning prior parameters in Bayesian decision tree learning,
and extend our results to decision tree regression as well as classical
pruning algorithms including min-cost complexity pruning. We introduce a
data-driven approach for optimizing the explainability versus accuracy
trade-off using decision trees, and demonstrate the empirical significance
of our approach.

Based on joint work with Nina Balcan, recognized with an Outstanding
student paper award to UAI 2024.

*Bio*: Dravyansh (Dravy) Sharma is a postdoctoral scholar supported by the
IDEAL fellowship under the mentorship of Avrim Blum and Aravindan
Srinivasan. He completed his PhD in the Computer Science Department at CMU,
advised by Nina Balcan. His research interests include machine learning
theory and algorithms, focusing on provable hyperparameter tuning,
adversarial robustness, and beyond worst-case analysis of algorithms. His
recent work develops techniques for tuning fundamental machine learning
algorithms to domain-specific data and introduces new, powerful robust
learning guarantees. His work has been published at top ML venues,
including NeurIPS, ICML, COLT, JMLR, AISTATS, UAI and AAAI. He has multiple
papers awarded with Oral presentations, has received the UAI 2024
Outstanding Student Paper Award, and has interned with Google Research and
Microsoft Research.



*In person: GHC 6115Zoom Link:
 https://cmu.zoom.us/j/99510233317?pwd=ZGx4aExNZ1FNaGY4SHI3Qlh0YjNWUT09
<https://cmu.zoom.us/j/99510233317?pwd=ZGx4aExNZ1FNaGY4SHI3Qlh0YjNWUT09>*


-Victor
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20241101/44f92534/attachment.html>