TRPO (Trust Region Policy Optimization) guarantees monotonic policy improvement using KL constraints, excelling in large neural network policy training for robotics and Atari games.

John Schulman, Sergey Levine, Philipp Moritz et al.

PESC outperforms EI methods in Bayesian optimization with unknown constraints.

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

Proposes a multi-layer LSTM encoder-decoder framework for unsupervised video representation learning, improving action recognition especially with limited labeled data.

Nitish Srivastava, Elman Mansimov, Ruslan Salakhutdinov

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

Berk Ustun, Cynthia Rudin

MADE uses masked autoencoders to enforce autoregressive constraints, boosting distribution estimation efficiency.

Mathieu Germain, Karol Gregor, Iain Murray et al.

Proposes Batch Normalization (BN) to standardize layer inputs, accelerating training and reducing internal covariate shift, achieving 4.9% top-5 error on ImageNet.

Sergey Ioffe, Christian Szegedy

Introduced a visual attention-based image captioning model achieving state-of-the-art performance on MS COCO.

Kelvin Xu, Jimmy Ba, Ryan Kiros et al.

Parallel-PC algorithm leverages multi-core processing to accelerate high-dimensional causal discovery, reducing runtime by 2-4x on gene expression datasets.

Thuc Duy Le, Tao Hoang, Jiuyong Li et al.

ORB-SLAM uses ORB features for real-time monocular SLAM, achieving high robustness and accuracy.

Raul Mur-Artal, J. M. M. Montiel, Juan D. Tardos

Proposed reciprocal scoring-based online dating recommendation leveraging multi-dimensional similarity measures, improving matching accuracy and user satisfaction.

Peng Xia, Benyuan Liu, Yizhou Sun et al.

Incorporating EXAA day-ahead prices as external regressors improves short-term European electricity price forecasts by 15-20%.

Florian Ziel, Rick Steinert, Sven Husmann

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.

Deep Fried ConvNets replace fully connected layers with adaptive Fastfood transforms, reducing parameters by over 90% without accuracy loss.

Zichao Yang, Marcin Moczulski, Misha Denil et al.

Adam combines first and second moment estimates for adaptive learning rates, accelerating large-scale stochastic optimization.

Diederik P. Kingma, Jimmy Ba

Elastic Averaging SGD (EASGD) leverages elastic force to enable efficient distributed deep learning, reducing communication overhead while enhancing exploration, leading to faster convergence and better accuracy.

Sixin Zhang, Anna Choromanska, Yann LeCun

Implicit regularization, not network size, controls capacity in deep learning.

Behnam Neyshabur, Ryota Tomioka, Nathan Srebro

By adding a cross-covariance penalty to autoencoders, hidden factors of variation in deep networks are discovered.

Brian Cheung, Jesse A. Livezey, Arjun K. Bansal et al.

MatConvNet is a MATLAB-based CNN toolbox supporting fast prototyping and GPU acceleration, enabling training on large datasets like ImageNet.

Andrea Vedaldi, Karel Lenc

This paper presents a high-level noise simulation model for CCD and CMOS sensors, validated for accuracy.

Mikhail Konnik, James Welsh

A spherical spin-glass mapping explains why SGD in large networks reaches low-energy, high-quality local minima.

Anna Choromanska, Mikael Henaff, Michael Mathieu et al.