What can transformers learn in-context? a case study of simple function classes

S Garg, D Tsipras, PS Liang… - Advances in Neural …, 2022‏ - proceedings.neurips.cc
In-context learning is the ability of a model to condition on a prompt sequence consisting of
in-context examples (input-output pairs corresponding to some task) along with a new query …

The staircase property: How hierarchical structure can guide deep learning

E Abbe, E Boix-Adsera, MS Brennan… - Advances in …, 2021‏ - proceedings.neurips.cc
This paper identifies a structural property of data distributions that enables deep neural
networks to learn hierarchically. We define the``staircase''property for functions over the …

Computational complexity of learning neural networks: Smoothness and degeneracy

A Daniely, N Srebro, G Vardi - Advances in Neural …, 2023‏ - proceedings.neurips.cc
Understanding when neural networks can be learned efficientlyis a fundamental question in
learning theory. Existing hardness results suggest that assumptions on both the input …

Connecting interpretability and robustness in decision trees through separation

M Moshkovitz, YY Yang… - … Conference on Machine …, 2021‏ - proceedings.mlr.press
Recent research has recognized interpretability and robustness as essential properties of
trustworthy classification. Curiously, a connection between robustness and interpretability …

Top-down induction of decision trees: rigorous guarantees and inherent limitations

G Blanc, J Lange, LY Tan - arxiv preprint arxiv:1911.07375, 2019‏ - arxiv.org
Consider the following heuristic for building a decision tree for a function $ f:\{0, 1\}^ n\to\{\pm
1\} $. Place the most influential variable $ x_i $ of $ f $ at the root, and recurse on the …

Harnessing the power of choices in decision tree learning

G Blanc, J Lange, C Pabbaraju… - Advances in …, 2023‏ - proceedings.neurips.cc
We propose a simple generalization of standard and empirically successful decision tree
learning algorithms such as ID3, C4. 5, and CART. These algorithms, which have been …

Provable guarantees for decision tree induction: the agnostic setting

G Blanc, J Lange, LY Tan - International Conference on …, 2020‏ - proceedings.mlr.press
We give strengthened provable guarantees on the performance of widely employed and
empirically successful {\sl top-down decision tree learning heuristics}. While prior works …

Intelligent Heuristics Are the Future of Computing

SH Teng - ACM Transactions on Intelligent Systems and …, 2023‏ - dl.acm.org
Back in 1988, the partial game trees explored by computer chess programs were among the
largest search structures in real-world computing. Because the game tree is too large to be …

The implications of local correlation on learning some deep functions

E Malach, S Shalev-Shwartz - Advances in Neural …, 2020‏ - proceedings.neurips.cc
It is known that learning deep neural-networks is computationally hard in the worst-case. In
fact, the proofs of such hardness results show that even weakly learning deep networks is …

Decision tree heuristics can fail, even in the smoothed setting

G Blanc, J Lange, M Qiao, LY Tan - arxiv preprint arxiv:2107.00819, 2021‏ - arxiv.org
Greedy decision tree learning heuristics are mainstays of machine learning practice, but
theoretical justification for their empirical success remains elusive. In fact, it has long been …