Improved stability and generalization for blackbox learned optimizers.
problem Stability and generalization issues in blackbox learned optimizers.
method Investigation using dynamical systems, modifications to optimizer architecture and meta-training procedure.
result Improved stability and generalization of learned optimizers.
Learned optimizers trained on many tasks improve model training efficiency and generalization.
problem Training models efficiently and effectively with minimal user intervention.
method Developed a neural network-based hierarchical optimizer trained on thousands of tasks.
result The learned optimizers generalize better to unseen tasks and exhibit unique behaviors.
Paper improves adversarial training using a learned optimizer.
problem Improving robustness of deep learning models against adversarial attacks.
method Empirically identified PGD attack's limitations and used a learning-to-learn framework to train an adaptive inner optimizer.
result The proposed framework consistently improves model robustness over traditional adversarial training methods.
VeLO learns versatile optimizers from deep learning tasks.
problem Training deep learning models with hand-designed optimizers.
method Meta-training a neural network optimizer on a wide variety of optimization tasks.
result The learned optimizer automatically adapts to different optimization tasks without hyperparameter tuning.
Learning-rate schedules for large models match optimization theory closely, leading to better training.
problem Improving training of large models with optimal learning rates.
method Used a bound from non-smooth convex optimization theory to match learning-rate schedules with practical benefits.
result Extending the learning-rate schedule with optimal learning-rate and transferring it across schedules improves model training.
Unified Bayesian Optimization framework for model selection balancing effectiveness and training efficiency.
problem Balancing model effectiveness and training efficiency in machine learning model selection.
method Proposes a unified Bayesian Optimization framework to jointly optimize model effectiveness and training efficiency.
result Models selected using the proposed framework significantly improve training efficiency while maintaining strong effectiveness.
Paper develops an efficient approach to reduce HPO time.
problem Challenges in determining optimal hyperparameters due to large number and training time.
method Nested Latin hypercube design for initialization, truncated additive Gaussian process model for calibration, sequential model-based algorithm for optimization.
result Demonstrates competitive performance on various machine learning models.
Adversarial training achieves optimal test error for shallow networks.
problem Achieving optimal adversarial test error for general data distributions.
method Applying new Rademacher complexity bounds and properties of optimal adversarial predictors.
result Adversarial training can achieve optimal adversarial test error for general data distributions.
Deep learning has shown that learned functions can dramatically outperform hand-designed functions on perceptual tasks. Analogously, this suggests that learned optimizers may similarly outperform current hand-designed optimizers, especially for specific problems. However, learned optimizers are notoriously difficult to…
Free adversarial training reduces the generalization gap compared to vanilla method.
problem Improving generalization in adversarial training.
method Analysis of algorithmic stability in free adversarial training.
result Free adversarial training shows a lower generalization gap.
Novel approach trains ASR models with less supervision using bilevel optimization.
problem Training acoustic models for ASR with minimal supervision.
method Bilevel optimization with unsupervised and supervised losses.
result Achieves superior performance compared to existing methods.
GT-DDP optimizer trains residual networks using game theory.
problem Applying DDP to residual networks.
method Game-theoretic DDP optimizer for residual networks.
result Improved training convergence and variance reduction.
Optimized biopharmaceutical seed train design reduces variability and saves time.
problem Designing robust biotechnological processes with cell cultures.
method Coupling uncertainty-based upstream simulation and Bayes optimization using Gaussian processes.
result Optimized seed train design results in lower cell density variability and reduced process duration.
GP-TS optimizes TLM pre-training hyperparameters efficiently.
problem Resource inefficiency in TLM pre-training.
method Bayesian optimization with Thompson sampling and Gaussian process.
result GP-TS achieves lower MLM loss in fewer epochs.
Paper optimizes training data distribution for better model performance across various deployment conditions.
problem Improving model accuracy when deployed with parameters far from training data.
method Developed adaptive algorithms based on bilevel or alternating optimization in the space of probability measures.
result Optimized training distributions lead to models with improved sample complexity and robustness to distribution shift.
Learning to Optimize is a recently proposed framework for learning optimization algorithms using reinforcement learning. In this paper, we explore learning an optimization algorithm for training shallow neural nets. Such high-dimensional stochastic optimization problems present interesting challenges for existing reinf…
Optimizes deep learning training by treating it as an optimal control problem.
problem Fragility of deep neural networks to adversarial inputs.
method Formulates adversarial training as a min-max optimization problem and interprets it as an optimal control problem.
result Provides the first convergence analysis of adversarial training algorithm.
We use GG distributions to optimize LLMs, reducing size and training time.
problem Lack of understanding in the statistical structure of LLMs.
method GG-based initialization, ACT, GCT.
result Smaller, faster models with minimal communication overhead.
A scalable method for training prediction models in predict-then-optimize
problem Training prediction models in the predict-then-optimize paradigm
method Decision-focused learning pipeline
result Decision quality competitive with state-of-the-art methods while reducing training time
New method trains quantized neural networks to global optimality.
problem Training optimal quantized neural networks is intractable due to combinatorial non-convex optimization.
method Convex optimization strategy using hidden convexity, semidefinite lifting, and Grothendieck's identity.
result Quantized NN problems can be solved to global optimality in polynomial-time.
Paper proposes a new method for training nonconvex models.
problem Training nonconvex models like neural networks.
method Successive functional gradient optimization using mirror descent in a function space.
result The method leads to better performance than standard training techniques.
Recently, neural networks trained as optimizers under the "learning to learn" or meta-learning framework have been shown to be effective for a broad range of optimization tasks including derivative-free black-box function optimization. Recurrent neural networks (RNNs) trained to optimize a diverse set of synthetic non-…
Optimal DP model training with public data improves privacy and accuracy.
problem Ensuring privacy while training models with public data.
method Proves optimal error rates for DP model training with public data, develops novel algorithms.
result Optimal error rates can be achieved by using public data or optimal DP algorithms.
This work improves diffusion models by estimating the optimal loss value for better training diagnostics.
problem The optimal loss value of diffusion models is unknown and not indicative of absolute data-fitting quality.
method Derive the optimal loss in closed form and develop effective estimators, including a stochastic variant.
result Unlocking the optimal loss as a metric for diagnosing training quality of diffusion models.
Paper tackles robustness in adversarial noise with a meta-optimizer.
problem Sensitivity to adversarial noise hinders machine learning deployment.
method Meta-optimizer learns to robustly optimize models using adversarial examples.
result Meta-optimizer transfers adversarial knowledge to new models without generating new examples.
Gradient-flow optimization is reinterpreted as a statistical inference problem.
problem Optimizing training duration and assessing model performance in deep learning.
method Develops a statistical framework for gradient-flow training, treating it as a random-effects model.
result Establishes asymptotic optimality for prediction and reduces reliance on validation splits.
Nonlinear conjugate gradient (NLCG) based optimizers have shown superior loss convergence properties compared to gradient descent based optimizers for traditional optimization problems. However, in Deep Neural Network (DNN) training, the dominant optimization algorithm of choice is still Stochastic Gradient Descent (SG…
Deriving and applying Proximal Policy Optimization to GFlowNets for efficient training of discrete sampling policies
problem Training stochastic policies to sample from structured discrete probability distributions
method Deriving policy gradient algorithms for GFlowNets and applying Proximal Policy Optimization
result Improved convergence speed and data efficiency compared to standard GFlowNet training objectives
ESPO optimizes LLMs for complex tasks by balancing fine-grained updates and stability.
problem Gradient underutilization in sequence-level optimization.
method Entropy Grouping Importance Sampling and Entropy Adaptive Clipping.
result ESPO accelerates convergence and achieves state-of-the-art performance.
This work analyzes and optimizes memory and compute costs of learned optimizers.
problem High memory and compute costs of learned optimizers.
method Identified and quantified design features of learned and hand-designed optimizers, constructed a more efficient learned optimizer.
result A learned optimizer that is faster and more memory efficient than previous work.
Transformers learn a mesa-optimizer to implement in-context learning.
problem Understanding the convergence of autoregressive training to a mesa-optimizer.
method Investigated a one-layer linear causal self-attention model autoregressively trained by gradient flow.
result Proved that autoregressive training converges to a gradient descent step for an OLS problem, validating the mesa-optimizer hypothesis.
We learn a compact surrogate model for optimization problems to reduce training and inference time.
problem Solving optimization problems with unknown parameters is computationally expensive and may lead to suboptimal solutions.
method We represent the optimization problem in terms of meta-variables and learn a low-dimensional surrogate model end-to-end with the predictive model.
result We achieve a large reduction in training and inference time, and improved performance.
MGD improves ML training efficiency and accuracy.
problem Optimizing large-scale ML model training configurations.
method Metagradient descent and smooth model training.
result Significantly improved model performance and learning rate schedules.
A new scaling law predicts optimal batch size for training models.
problem Finding the optimal batch size for training models efficiently.
method Proposed a three-term scaling law that considers model size, training data, training steps, and batch size.
result The three-term law accurately recovers the optimal batch size and can be robustly fit with fewer training runs.
AskewSGD optimizes quantized neural networks with interval-constrained optimization.
problem Training deep neural networks with quantized weights.
method Formulates QNN training as smoothed interval-constrained optimization, proposes AskewSGD for solving each subproblem.
result AskewSGD avoids projections and allows infeasible iterates, performs better than state-of-the-art methods.
State-level minimum Bayes risk (sMBR) training has become the de facto standard for sequence-level training of speech recognition acoustic models. It has an elegant formulation using the expectation semiring, and gives large improvements in word error rate (WER) over models trained solely using cross-entropy (CE) or co…
New insights into GNN optimization reveal skip connections and depth accelerate training.
problem Understanding and optimizing the training of Graph Neural Networks (GNNs).
method Analysis of gradient dynamics in linearized GNNs and empirical validation.
result GNNs are implicitly accelerated by skip connections, more depth, and good label distribution during training.
PLUMAGE improves large model training efficiency and stability.
problem Accelerator memory and networking constraints during large model training.
method Probabilistic Low rank Unbiased Minimum Variance Gradient Estimator (PLUMAGE) that resolves bias and variance issues.
result PLUMAGE reduces training loss by 28% on average across the GLUE benchmark.
Efficient algorithms speed up adversarial training for linear models.
problem Adversarial training for linear models is computationally expensive.
method Tailored optimization algorithms for regression and classification.
result Significantly faster convergence rates for large-scale problems.
New framework trains Schrödinger Bridge models using SDEs for generative tasks.
problem Unclear relation between SB optimization and modern generative model training.
method Forward-Backward SDEs theory for likelihood training of SB models.
result Training algorithm achieves comparable results on image generation datasets.
This study explains RL training dynamics in LLMs, focusing on token-level optimization and reasoning pattern reshaping.
problem Understanding the training dynamics of RL in LLMs to improve their reasoning capabilities.
method Empirical analysis and theoretical modeling of RL training process, focusing on reasoning patterns and token optimization.
result RL primarily optimizes a sparse subset of critical tokens, reshaping reasoning pattern distributions and affecting model performance.
GFM models neural network training as a dynamical system to forecast final weights.
problem Computational intensity and inefficiency in training deep neural networks.
method Gradient Flow Matching (GFM) treats training as a dynamical system with learned vector fields.
result GFM achieves forecasting accuracy competitive with Transformer-based models and significantly outperforms classical baselines.
Direct optimization of binary latent VAEs achieves competitive results without sampling.
problem Training VAEs with discrete latent variables using standard methods is challenging.
method Applied evolutionary algorithms to directly optimize discrete latent distributions.
result Direct optimization is efficient and competitive in zero-shot learning.
Automated Budget Constrained Training optimizes model training under time constraints.
problem Balancing model quality and computational cost in constrained time.
method Developed a hyperparameter optimisation algorithm that learns the relationship between hyperparameters, model quality, and computational cost.
result The algorithm optimally decides whether to terminate or continue training, and what hyperparameters to use.
Given a sequential learning algorithm and a target model, sequential machine teaching aims to find the shortest training sequence to drive the learning algorithm to the target model. We present the first principled way to find such shortest training sequences. Our key insight is to formulate sequential machine teaching…
New methods optimize training VQAs without barren plateaus, improving efficiency and applicability.
problem Barren plateaus in training variational quantum algorithms.
method Derive adaptive learning rates and use Gaussian kernels to optimize movement in parameter space.
result Optimized training methods outperform other routines and can train VQAs free of barren plateaus.
State-of-the art vision models can achieve superhuman performance on image classification tasks when testing and training data come from the same distribution. However, when models are tested on corrupted images (e.g. due to scale changes, translations, or shifts in brightness or contrast), performance degrades signifi…
Optimizing data movement significantly improves transformer training efficiency.
problem Data movement is a major bottleneck in training transformers.
method Developed a recipe to globally optimize data movement in transformers.
result Achieved up to 1.30x performance improvement over state-of-the-art frameworks.