PGS uses neural networks to improve policies online without search trees.
problem Limited scalability of Monte Carlo Tree Search (MCTS) for high branching factor games.
method Adapts a neural network simulation policy via policy gradient updates, avoiding search trees.
result PGS achieves comparable performance to MCTS and defeats strong Hex agents.
Trust-region methods and natural gradients are equivalent in certain policy search scenarios.
problem Improving policy search methods in continuous control tasks.
method Introducing compatible policy search (COPOS) that uses natural parameterization and compatible value function approximation to control entropy loss.
result COPOS yields state-of-the-art results in challenging tasks and reduces entropy loss.
Probabilistic line search improves stochastic optimization efficiency.
problem Lack of direct line search methods for stochastic optimization.
method Combines deterministic line search structure with Bayesian optimization concepts.
result Effective removal of learning rate definition for SGD.
DARTS efficiently searches for high-performance architectures using gradient descent.
problem Scalability challenge of architecture search in machine learning.
method Differentiable relaxation of architecture representation for continuous search space.
result Orders of magnitude faster than non-differentiable techniques.
SNAS efficiently searches neural architectures using stochastic optimization.
problem Efficiently searching for optimal neural architectures.
method SNAS trains parameters of both neural operations and architecture distribution in a single round of backpropagation, using a novel search gradient and locally decomposable rewards.
result SNAS achieves state-of-the-art accuracy with fewer training epochs compared to other NAS methods.
Paper stabilizes DARTS algorithm for better neural architecture search.
problem Weak stability of DARTS algorithm leading to unreliable results.
method Amended gradient estimation method to bridge optimization gap.
result Significant improvement in search stability and larger search spaces explored.
NASP uses proximal gradient descent to speed up neural architecture search.
problem Efficiently search for high-performance neural architectures.
method Differentiable Neural Architecture Search using Proximal gradient descent.
result NASP achieves 10 times speedup over DARTS while maintaining high performance.
New stochastic gradient descent with random search directions improves efficiency and convergence.
problem Efficiency and convergence of stochastic gradient descent methods.
method Developed a new class of stochastic gradient descent algorithms with random search directions.
result Established almost sure convergence and provided Lp rates of convergence. Automated meta-learning improves model performance on small datasets.
problem Improving machine learning model performance on limited data.
method Gradient-based meta-learning combined with automated neural architecture search.
result Automatically found meta-learner achieved 74.65% accuracy on 5-shot 5-way Mini-ImageNet, 11.54% better than MAML.
Guided Evolutionary Strategies uses surrogate gradients to improve optimization.
problem Optimizing functions with unknown true gradients but available surrogate gradients.
method Combines random search with a search distribution elongated along surrogate gradient directions.
result Improves optimization performance over standard evolutionary strategies and first-order methods.
Paper proposes MCTSPO for better reinforcement learning policy optimization.
problem Local optima and saddle points in gradient-based methods and poor initialization in gradient-free methods.
method Monte-Carlo tree search combined with gradient-free optimization.
result Improved performance on reinforcement learning tasks with deceptive or sparse reward functions.
New method finds optimal learning rates for neural nets.
problem Finding optimal learning rates in stochastic neural networks.
method Gradient-only line searches using Non-negative Associative Gradient Projection Points (NN-GPPs).
result Learning rates can be reliably resolved as step sizes along search directions.
Novel BSG method for efficient stochastic optimization.
problem Efficient optimization of non-convex surfaces in stochastic settings.
method Binary search combined with first order gradient optimization.
result BSG produces more promising results and better generalization than other methods.
Develops a robust, fast, and widely-applicable neural architecture search method.
problem Inability of current NAS methods to be easily applied to new problems.
method Adaptive stochastic natural gradient method for simultaneous optimization of weights and architecture.
result Near state-of-the-art performances with low computational budgets.
Gradient-based feature selection for large datasets.
problem Feature selection for large datasets with high-order correlations.
method Iterative mini-batch calculation, discrete-to-continuous relaxation.
result Efficiently finds higher-order feature correlations in both N > D and N < D regimes.
Improved SGD with line-search achieves fast convergence rates for various models.
problem Achieving fast convergence rates for stochastic gradient descent (SGD) in over-parameterized models.
method Proposes using line-search techniques to automatically set the step-size in SGD, proving convergence rates for convex, strongly-convex, and non-convex functions.
result SGD with Armijo line-search attains deterministic convergence rates for convex and strongly-convex functions, and linear convergence for non-convex functions.
Paper tackles NAS problem by modeling it as a sparse supernet.
problem Neural Architecture Search (NAS) problem, particularly Mixed-Path Search.
method Model NAS as a sparse supernet with sparsity constraints. Use hierarchical accelerated proximal gradient algorithm for optimization.
result Proposed method finds compact, general, and powerful neural architectures.
In deterministic optimization, line searches are a standard tool ensuring stability and efficiency. Where only stochastic gradients are available, no direct equivalent has so far been formulated, because uncertain gradients do not allow for a strict sequence of decisions collapsing the search space. We construct a prob…
A new Randomized-Hyperopt method improves XGBoost hyperparameter tuning.
problem Improving the performance of XGBoost through hyperparameter optimization.
method Proposes Randomized-Hyperopt for XGBoost hyperparameter tuning.
result Randomized-Hyperopt outperforms other methods in terms of accuracy and execution time.
A new method combines extrapolation and line search for solving nonconvex, nonsmooth optimization problems.
problem Nonconvex, nonsmooth optimization problems in machine learning and image processing.
method Proximal gradient method with extrapolation and line search (PGels).
result The method reduces to existing algorithms under proper parameter choices and converges to stationary points.
New method reduces NAS search time and complexity.
problem High computational cost and complexity in NAS.
method Differentiable search space with annealing and pruning.
result Achieves 1.68% error on CIFAR-10 with 0.2 GPU days.
New algorithm optimizes AUC in binary classification and changepoint detection.
problem Difficult to optimize AUC in binary classification and changepoint detection.
method Proposes efficient path-following algorithms for choosing optimal learning rate.
result Proposed line search algorithm computes complete AUM/AUC representation.
Paper uses CMAB to improve NAS efficiency and accuracy.
problem Improving efficiency and accuracy of NAS for DNNs.
method Formulated NAS as CMAB, used Nested Monte-Carlo Search.
result Discovered cell structure achieves comparable accuracy to state-of-the-art, 20x faster.
This paper improves search efficiency by augmenting autoencoder encodings.
problem Improving search speed and relevance in information retrieval.
method Gradient Augmented Information Retrieval with Autoencoders and Semantic Hashing.
result Gradient Augmented Search (GSA) enhances TF-IDF-based systems.
Armijo line-search speeds up gradient descent for various functions.
problem Improving convergence rate of gradient descent.
method Applying Armijo line-search to adjust step-size in gradient descent.
result GD with Armijo line-search converges faster than GD with a fixed step-size.
Stochastic gradient descent on manifolds improves low-rank approximation.
problem Efficiently approximate large matrices with lower rank.
method Stochastic gradient descent on a manifold.
result Algorithm outperforms Euclidean space methods on Netflix Prize data.
Adaptive gradient methods converge faster with over-parameterization and line-search.
problem Training over-parameterized models using adaptive gradient methods.
method Simplified setting of smooth, convex losses with over-parameterized models, proving convergence rates and demonstrating improvements with line-search techniques.
result Adaptive gradient methods, particularly AMSGrad, converge faster with line-search techniques.
A new scheme reduces global search cost by a square root factor.
problem Challenges in finding global minimum of cost functions.
method Gradient descent combined with a biased crossover of two good solutions.
result Quadratic speedup of global search efficiency.
NVA combines variational posteriors, annealing, and natural-gradient learning for multimodal optimization.
problem Finding multiple global and local modes in nonconvex objectives.
method NVA integrates variational posteriors, annealing, and natural-gradient learning.
result NVA outperforms gradient descent and evolution strategies on simulations and real-world problems.
Petridish efficiently searches neural architectures by iteratively adding shortcut connections.
problem Finding efficient neural architectures for various tasks.
method Iteratively adds shortcut connections to existing network layers, motivated by feature selection.
result Petridish efficiently finds competitive models with few GPU days.
A neural network approach to Monte-Carlo tree search.
problem Improving tree search algorithms for planning problems.
method Learning neural network architecture to control search parameters.
result The learned search algorithm outperformed traditional MCTS.
RMGD uses bandit theory to optimize mini-batch size for faster and better performance.
problem Determining the optimal mini-batch size for gradient descent is time-consuming.
method Resilient Mini-batch Gradient Descent (RMGD) using Multi-Armed Bandit.
result RMGD achieves better performance than grid search in less time.
Stochastic quasi-Newton tackles noisy gradients in optimization.
problem Optimizing with noisy data in stochastic settings.
method Extends quasi-Newton methods to handle stochastic gradients through flexible Hessian modeling and line-search regularization.
result Demonstrates superior performance in maximum likelihood estimation for complex models.
Differentially-private FNAS protects privacy while collaboratively searching for neural architectures.
problem Collaborative neural architecture search with privacy concerns.
method Federated Neural Architecture Search (FNAS) with differential privacy (DP-FNAS).
result DP-FNAS can search for highly-performant neural architectures while protecting individual parties' privacy.
Bayesian optimization improves policy search in reinforcement learning.
problem Finding optimal policies with high variance estimates from random samples.
method Develops an algorithm combining Bayesian optimization and policy gradients.
result Improves sample complexity and reduces variance in empirical evaluations.
Paper introduces a privacy-preserving line search method for optimization.
problem Optimization performance depends on step size tuning, which is difficult and privacy-sensitive.
method Introduces a stochastic adaptive line search algorithm that satisfies differential privacy.
result The algorithm efficiently uses privacy budget and outperforms existing private optimizers.
Gradients help find global optima in complex functions.
problem Finding global optima in functions with many local minima.
method A principle for generating search directions from non-local quadratic approximants based on gradients.
result The proposed algorithm and CMA-ES perform better than random reinitialized BFGS.
Unified framework for gradient-free MDS improves efficiency and accuracy.
problem Efficiently solving Multidimensional Scaling problems without derivatives.
method Bootstrapped Coordinate Search (BS CSMDS) for MDS, using a probability matrix to guide search.
result BS CSMDS achieves significant speedup and maintains error rate compared to other CSMDS methods.
New algorithms improve contextual search in the presence of adversarial corruptions.
problem Improving search accuracy in dynamic pricing settings with corrupted responses.
method Two algorithms based on binary search and gradient descent methods.
result Achieve near-optimal regret in the absence of adversarial corruptions and gracefully degrade with corrupted agents.
SALSA automatically adjusts learning rates in stochastic gradient methods.
problem Automatic adjustment of learning rates in stochastic gradient methods.
method SALSA uses a line-search procedure to gradually increase the learning rate, then a statistical test to decrease it.
result SALSA matches the performance of best hand-tuned learning rate schedules in deep learning tasks.
Two new algorithms improve neural architecture search efficiency.
problem Optimizing neural architecture search for faster and more accurate models.
method Introduces NASGD and NASAGD using accelerated gradient descent on a semi-discrete space.
result Achieves comparable accuracy with 40x fewer architectures in 12 hours.
DrNAS improves neural architecture search with Dirichlet distribution and progressive learning.
problem Efficiently search for neural architectures with improved generalization and exploration.
method Formulates architecture search as a distribution learning problem using Dirichlet distribution and gradient-based optimization. Introduces a progressive learning scheme to handle large-scale tasks.
result Achieves state-of-the-art results on CIFAR-10 and ImageNet, demonstrating improved generalization and exploration.
A new method optimizes neural sequence models for better task performance.
problem Training neural sequence models with maximum likelihood estimation ignores task losses.
method Maximum likelihood guided parameter search (MGS) in the parameter space.
result MGS optimizes sequence-level losses, reducing repetition and non-termination.
New algorithms improve neural architecture search with faster convergence.
problem Improving efficiency and accuracy of neural architecture search.
method Geometry-aware gradient algorithms to optimize continuous relaxation of discrete search spaces.
result Exceeds state-of-the-art results on CIFAR and ImageNet benchmarks.
CAGES optimizes expensive RL problems by efficiently learning gradients from multiple sources.
problem Optimizing expensive-to-evaluate functions in high-dimensional spaces.
method Cost-Aware Gradient Entropy Search (CAGES) for multi-fidelity Bayesian optimization.
result Significant performance improvements on synthetic and RL benchmark problems.
GOLS finds activation functions affect training robustness, especially ReLU.
problem Investigate how different activation functions impact GOLS in neural network training.
method Identify SNN-GPPs for GOLS, analyze activation function effects on gradient continuity.
result GOLS robust for most activation functions but sensitive to ReLU.
PIPPS solves deep learning's exploding gradient problem by reparameterization gradients.
problem Exploding gradients in deep learning and model-based RL.
method Develops PIPPS framework, a flexible policy search method robust to chaos-like gradients.
result PIPPS improves over reparameterization gradients by up to 10^6 times.
GOLS-I automatically determines learning rates for various neural network training algorithms.
problem Adapting learning rates in stochastic training algorithms for neural networks.
method Gradient-Only Line Search (GOLS-I) for automatically setting learning rates.
result GOLS-I learning rate schedules are competitive with manually tuned rates across multiple algorithms, architectures, datasets, and loss functions.