Proposes a scalable variational information maximization algorithm for high-dimensional visual inputs, enhancing intrinsic motivation in reinforcement learning.

Shakir Mohamed, Danilo Jimenez Rezende

Dynamic Matrix Factorization with Priors improves sparse recommendation, reaching 0.9276 AUC on AmazonMovies.

Robin Devooght, Nicolas Kourtellis, Amin Mantrach

Pointer Networks leverage attention as pointers, enabling variable-length input-output mapping, successfully applied to convex hull, Delaunay triangulation, and TSP.

Oriol Vinyals, Meire Fortunato, Navdeep Jaitly

Kernel Manifold Alignment (KEMA) enables multi-source, unpaired domain alignment with superior performance on synthetic and real datasets.

Devis Tuia, Gustau Camps-Valls

Knowledge distillation transfers ensemble model knowledge into a single small model, achieving near-ensemble performance on MNIST and speech recognition tasks.

Geoffrey Hinton, Oriol Vinyals, Jeff Dean

The Bayesian Case Model (BCM) integrates case-based reasoning with a generative framework.

Been Kim, Cynthia Rudin, Julie Shah

PESC outperforms EI methods in Bayesian optimization with unknown constraints.

José Miguel Hernández-Lobato, Michael A. Gelbart, Matthew W. Hoffman et al.

SLIM method optimizes medical scoring systems' accuracy and sparsity via integer programming.

Berk Ustun, Cynthia Rudin

Introducing Quasi-Monte Carlo feature maps for shift-invariant kernels, reducing feature dimension and improving approximation accuracy.

Haim Avron, Vikas Sindhwani, Jiyan Yang et al.

Graph-Sparse LDA uses ontology-guided sparsity, compressing an ASD topic from 119 diagnoses to 6 concepts without losing predictive performance.

Finale Doshi-Velez, Byron Wallace, Ryan Adams

Proposes a gradient reversal-based unsupervised domain adaptation method, significantly improving cross-domain image classification accuracy.

Yaroslav Ganin, Victor Lempitsky

Introduces Predictive Entropy Search (PES) for efficient black-box function optimization.

José Miguel Hernández-Lobato, Matthew W. Hoffman, Zoubin Ghahramani

Proposes zero-bias autoencoders with decoupled activation functions, enabling high-dimensional data representation without extra regularization, improving recognition accuracy.

Kishore Konda, Roland Memisevic, David Krueger

Nested Dropout enforces ordered representations, ensuring unit identifiability, equivalent to PCA, enabling fast large-scale retrieval and adaptive compression.

Oren Rippel, Michael A. Gelbart, Ryan P. Adams

Petuum employs error-tolerant bounded synchronization and dynamic scheduling to enable scalable distributed ML with billions of parameters.

Eric P. Xing, Qirong Ho, Wei Dai et al.

Dropout significantly reduces catastrophic forgetting in neural networks, with experimental data showing 15% better retention of old tasks and 20% improved adaptation to new tasks.

Ian J. Goodfellow, Mehdi Mirza, Da Xiao et al.

Introduces reparameterization and auto-encoding variational Bayes (AEVB) for efficient inference in continuous latent variable models.

Diederik P Kingma, Max Welling

Proposed an efficient training method for NADE models, enhancing density estimation performance.

Benigno Uria, Iain Murray, Hugo Larochelle

SLIM combines 0–1 loss, L0 sparsity, and integer constraints to produce accurate, hand-computable scoring systems.

Berk Ustun, Stefano Tracà, Cynthia Rudin

This paper applies inductive conformal prediction combined with projection and pseudo-density techniques to construct distribution-free, finite-sample guaranteed prediction bands and clustering trees for functional data, enabling outlier detection and structure exploration.

Jing Lei, Alessandro Rinaldo, Larry Wasserman