With 49 patents and over 100 inventions, Henry built an empire catering to women and children. Beulah Louise Henry invented everything from ice cream makers to radio dolls--despite a world that didn't take her seriously. Breakthroughs, discoveries, and DIY tips sent six days a week. Beulah Louise Henry was just nine years old when she came up with her first invention in 1896, a device that allowed a man to tip his hat without ever putting down his newspaper. By her death in 1973, at the age of 85, she'd come up with so many more--a doll with eyes that changed color with the press of a button, a sewing machine without a bobbin (a threaded spool that slowed down work because it had to be frequently refilled), a clock designed to help kids learn to tell time, and others--that the press even dubbed Henry "Lady Edison."

What are the right computational primitives for exploration?

The training of language models is typically not an end-to-end process.

Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/

The nineteenth century saw the emergence of proofs so involved that they could only be verified by a handful of specialists.

We give new differentially private algorithms for the classic problems of learning decision lists and large-margin halfspaces in the PAC and online models. In the PAC model, we give a computationally efficient algorithm for learning decision lists with minimal sample overhead over the best non-private algorithms. In the online model, we give a private analog of the influential Winnow algorithm for learning halfspaces with mistake bound polylogarithmic in the dimension and inverse polynomial in the margin. As an application, we describe how to privately learn decision lists in the online model, qualitatively matching state-of-the art non-private guarantees.

Contextual bandits are a core technology for personalized mobile health interventions, where decision-making requires adapting to complex, non-linear user behaviors. While Thompson Sampling (TS) is a preferred strategy for these problems, its performance hinges on the quality of the underlying reward model. Standard linear models suffer from high bias, while neural network approaches are often brittle and difficult to tune in online settings. Conversely, tree ensembles dominate tabular data prediction but typically rely on heuristic uncertainty quantification, lacking a principled probabilistic basis for TS. We propose Bayesian Forest Thompson Sampling (BFTS), the first contextual bandit algorithm to integrate Bayesian Additive Regression Trees (BART), a fully probabilistic sum-of-trees model, directly into the exploration loop. We prove that BFTS is theoretically sound, deriving an information-theoretic Bayesian regret bound of $\tilde{O}(\sqrt{T})$. As a complementary result, we establish frequentist minimax optimality for a "feel-good" variant, confirming the structural suitability of BART priors for non-parametric bandits. Empirically, BFTS achieves state-of-the-art regret on tabular benchmarks with near-nominal uncertainty calibration. Furthermore, in an offline policy evaluation on the Drink Less micro-randomized trial, BFTS improves engagement rates by over 30% compared to the deployed policy, demonstrating its practical effectiveness for behavioral interventions.

Work done while at IBMResearch - Almaden.