ICML 2018 accepted papers on Optimization
86 of the 623 papers accepted at ICML 2018 (13.8%) are on Optimization, compared with 13.8% at ICML 2017. They fall into 4 topics.
See ICML 2018 in the interactive map
Submitting to ICML? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Nonconvex and smooth optimization · 39 papers
- The Mechanics of n-Player Differentiable Games (138 citations)
- SGD and Hogwild! Convergence Without the Bounded Gradients Assumption (100 citations)
- Escaping Saddles with Stochastic Gradients (79 citations)
- Computational Optimal Transport: Complexity by Accelerated Gradient Descent Is Better Than by Sinkhorn’s Algorithm (66 citations)
- A Simple Stochastic Variance Reduced Algorithm with Fast Convergence Rates (58 citations)
- ADMM and Accelerated ADMM as Continuous Dynamical Systems (44 citations)
- On Acceleration with Noise-Corrupted Gradients (44 citations)
- Convergence guarantees for a class of non-convex and non-smooth optimization problems (43 citations)
- Stochastic Proximal Algorithms for AUC Maximization (38 citations)
- Fast Stochastic AUC Maximization with $O(1/n)$-Convergence Rate (35 citations)
- On the Theory of Variance Reduction for Stochastic Gradient Monte Carlo (26 citations)
- Riemannian Stochastic Recursive Gradient Algorithm (25 citations)
and 27 more in the interactive map.
Distributed and compressed · 16 papers
- Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates (612 citations)
- The Hidden Vulnerability of Distributed Learning in Byzantium (557 citations)
- Communication-Computation Efficient Gradient Coding (126 citations)
- signSGD: Compressed Optimisation for Non-Convex Problems (87 citations)
- Asynchronous Decentralized Parallel Stochastic Gradient Descent (66 citations)
- Improving the Privacy and Accuracy of ADMM-Based Distributed Algorithms (56 citations)
- Distributed Asynchronous Optimization with Unbounded Delays: How Slow Can You Go? (37 citations)
- Error Compensated Quantized SGD and its Applications to Large-scale Distributed Optimization (35 citations)
- Asynchronous Byzantine Machine Learning (the case of SGD) (25 citations)
- A Delay-tolerant Proximal-Gradient Algorithm for Distributed Learning (20 citations)
- Asynchronous Stochastic Quasi-Newton MCMC for Non-Convex Optimization (17 citations)
- Optimal Distributed Learning with Multi-pass Stochastic Gradient Methods (11 citations)
and 4 more in the interactive map.
Optimizers · 16 papers
- Learning to Reweight Examples for Robust Deep Learning (841 citations)
- Which Training Methods for GANs do actually Converge? (814 citations)
- Bilevel Programming for Hyperparameter Optimization and Meta-Learning (105 citations)
- Not All Samples Are Created Equal: Deep Learning with Importance Sampling (104 citations)
- Dissecting Adam: The Sign, Magnitude and Variance of Stochastic Gradients (78 citations)
- Gradient-Based Meta-Learning with Learned Layerwise Metric and Subspace (72 citations)
- Decoupled Parallel Backpropagation with Convergence Guarantee (60 citations)
- A Progressive Batching L-BFGS Method for Machine Learning (58 citations)
- Shampoo: Preconditioned Stochastic Tensor Optimization (52 citations)
- On the Limitations of First-Order Approximation in GAN Dynamics (43 citations)
- Stein Variational Gradient Descent Without Gradient (25 citations)
- Accelerating Natural Gradient with Higher-Order Invariance (6 citations)
and 4 more in the interactive map.
Implicit bias and generalization · 15 papers
- Essentially No Barriers in Neural Network Energy Landscape (178 citations)
- Characterizing Implicit Bias in Terms of Optimization Geometry (168 citations)
- Curriculum Learning by Transfer Learning: Theory and Experiments with Deep Networks (151 citations)
- On the Optimization of Deep Networks: Implicit Acceleration by Overparameterization (136 citations)
- The Power of Interpolation: Understanding the Effectiveness of SGD in Modern Over-parametrized Learning (121 citations)
- Implicit Regularization in Nonconvex Statistical Estimation: Gradient Descent Converges Linearly for Phase Retrieval and Matrix Completion (62 citations)
- Data-Dependent Stability of Stochastic Gradient Descent (60 citations)
- Gradient descent with identity initialization efficiently learns positive definite linear transformations by deep residual networks (48 citations)
- An Alternative View: When Does SGD Escape Local Minima? (38 citations)
- Understanding Generalization and Optimization Performance of Deep CNNs (26 citations)
- The Dynamics of Learning: A Random Matrix Approach (21 citations)
- Gradient Descent Learns One-hidden-layer CNN: Don’t be Afraid of Spurious Local Minima (20 citations)
and 3 more in the interactive map.
