We invite submissions with theoretical and algorithmic contributions to new or already existing learning problems including but not limited to:

Comparison of the strength of learning models and the design and evaluation of novel algorithms for learning problems in established learning-theoretic settings such as:

• Statistical learning theory

• Supervised learning and regression

• Statistical learning theory

• On-line learning

• Inductive inference

• Query models

• Unsupervised learning

• Clustering

• Semi-supervised and active learning

• Stochastic optimization

• High dimensional and non-parametric inference

• Exploration-exploitation tradeoff, bandit theory

• Reinforcement learning, planning, control

• Learning with additional constraints, e.g., communication, time or memory budget, or privacy

Analysis of the theoretical properties of existing algorithms such as:

• Boosting

• Kernel-based methods, SVM

• Bayesian methods

• Graph- and/or manifold-based methods

• Methods for latent-variable estimation and/or clustering

• Decision tree methods

• Information-based methods, MDL

• Neural networks

Analyses could include generalization, speed of convergence, computational complexity, or sample complexity.

Definition and analysis of new learning models. Models might identify and formalize classes of learning problems inadequately addressed by existing theory or capture salient properties of important concrete applications.

We are also interested in papers that include viewpoints that are new to the ALT community. We welcome experimental and algorithmic papers provided they are relevant to the focus of the conference by elucidating theoretical results, or by pointing out interesting and not well understood behavior that could stimulate theoretical analysis.