6 papers
Of Dice and Games: A Theory of Generalized Boosting
Marco Bressan, Nataly Brukhim, Nicolò Cesa-Bianchi +4
Cost-sensitive loss functions are crucial in many real-world prediction problems, where different types of errors are penalized differently; for example, in medical diagnosis, a fa…
Improved Regret Bounds for Bandits with Expert Advice
Nicolò Cesa-Bianchi, Khaled Eldowa, Emmanuel Esposito +1
In this research note, we revisit the bandits with expert advice problem. Under a restricted feedback model, we prove a lower bound of order for the worst-cas…
Efficient Algorithms for Learning Monophonic Halfspaces in Graphs
Marco Bressan, Emmanuel Esposito, Maximilian Thiessen
We study the problem of learning a binary classifier on the vertices of a graph. In particular, we consider classifiers given by monophonic halfspaces, partitions of the vertices t…
A Theory of Interpretable Approximations
Marco Bressan, Nicolò Cesa-Bianchi, Emmanuel Esposito +3
Can a deep neural network be approximated by a small decision tree based on simple features? This question and its variants are behind the growing demand for machine learning model…
An Improved Uniform Convergence Bound with Fat-Shattering Dimension
Roberto Colomboni, Emmanuel Esposito, Andrea Paudice
The fat-shattering dimension characterizes the uniform convergence property of real-valued functions. The state-of-the-art upper bounds feature a multiplicative squared logarithmic…
Delayed Bandits: When Do Intermediate Observations Help?
Emmanuel Esposito, Saeed Masoudian, Hao Qiu +3
We study a -armed bandit with delayed feedback and intermediate observations. We consider a model where intermediate observations have a form of a finite state, which is observe…