FisherSFT selects informative examples to fine-tune LLMs efficiently.
problem Adapting large language models to new domains efficiently.
method Selects examples maximizing information gain using Hessian of log-likelihood.
result Empirically demonstrates improved performance with reduced computational cost.
We study the problem of online path learning with non-additive gains, which is a central problem appearing in several applications, including ensemble structured prediction. We present new online algorithms for path learning with non-additive count-based gains for the three settings of full information, semi-bandit and…
Active inference selects actions to maximize information gain, aiding structure learning.
problem Learning the structure of underlying world models.
method Active inference selects actions based on expected free energy, which includes information gain and value.
result Actions that maximize information gain help disambiguate among alternative models.
Meta-active learning optimizes control of safety-critical systems by efficiently learning dynamics and configurations.
problem Efficiently learning system dynamics and optimal configurations for safety-critical systems like deep brain stimulation.
method Meta-learning an acquisition function using LSTM, cast as meta-learning, with a mixed-integer linear program policy.
result Achieved a 46% increase in information gain and a 20% speedup in computation time over baselines.
Study uses surrogate data to improve treatment effect estimation with scarce outcome data.
problem Limited outcome data hinders estimating treatment effects.
method Uses abundant surrogate data to estimate treatment effects without stringent assumptions.
result Improves precision of treatment effect estimation.
Develops methods to estimate gradient of EIG for Bayesian Experimental Design.
problem Optimizing Bayesian inference through efficient experimental design.
method Introduces UEEG-MCMC and BEEG-AP methods for estimating EIG gradient.
result Both methods improve upon existing benchmarks in EIG optimization.
We present a novel high frequency residual learning framework, which leads to a highly efficient multi-scale network (MSNet) architecture for mobile and embedded vision problems. The architecture utilizes two networks: a low resolution network to efficiently approximate low frequency components and a high resolution ne…
Meta-learning algorithm improves AI efficiency by teaching itself.
problem Overcoming meta-optimization challenges in AI learning.
method Bootstraps a target from the meta-learner and optimizes the meta-learner by minimizing distance under a chosen metric.
result Achieves state-of-the-art performance on Atari ALE benchmark and demonstrates efficiency gains in multi-task learning.
We propose a tree regularization framework, which enables many tree models to perform feature selection efficiently. The key idea of the regularization framework is to penalize selecting a new feature for splitting when its gain (e.g. information gain) is similar to the features used in previous splits. The regularizat…
ActiveCQ improves causal quantity estimation with active learning and Gaussian Processes.
problem Estimating causal quantities requires large datasets, which are costly.
method Unified framework using Gaussian Processes and conditional mean embeddings for distribution estimation. Derived principled acquisition strategies based on information gain and total variance reduction.
result Framework significantly outperforms baselines in sample efficiency across various causal quantities.
Bayesian optimal experimental design (BOED) is a principled framework for making efficient use of limited experimental resources. Unfortunately, its applicability is hampered by the difficulty of obtaining accurate estimates of the expected information gain (EIG) of an experiment. To address this, we introduce several …
We propose a novel information-theoretic approach for Bayesian optimization called Predictive Entropy Search (PES). At each iteration, PES selects the next evaluation point that maximizes the expected information gained with respect to the global maximum. PES codifies this intractable acquisition function in terms of t…
Logarithmic-time schedules boost large-scale language model training efficiency.
problem Improving performance and efficiency in large-scale language model training.
method Designing time-varying hyperparameters (β1,β2,λ) for AdamW, specifically logarithmic-time scheduling with damping mechanisms. result ADANA optimizer achieves up to 40% compute efficiency compared to tuned AdamW, with gains persisting as model scale increases.
Gradient surgery improves multi-task learning efficiency.
problem Challenges in sharing structure across multiple tasks.
method Gradient projection onto normal plane of conflicting gradients.
result Substantial gains in efficiency and performance.
Paper optimizes financial trading strategies under uncertain market conditions.
problem Guaranteeing robust positive expected profits in financial systems.
method Transformed semi-infinite constraints into structured policies and proposed a novel graphical approach.
result Demonstrated superior risk-adjusted returns and downside risk compared to conventional strategies.
New method improves robustness of Bayesian experimental design.
problem Bayesian experimental design's sensitivity to prior distribution changes.
method Introduces robust expected information gain (REIG) and uses KL-divergence ambiguity sets.
result REIG stabilizes sampling-based EIG estimation and compensates for prior variability.
Hardware accelerations of deep learning systems have been extensively investigated in industry and academia. The aim of this paper is to achieve ultra-high energy efficiency and performance for hardware implementations of deep neural networks (DNNs). An algorithm-hardware co-optimization framework is developed, which i…
Efficiently identifies key input variables for expensive functions using active learning.
problem Efficiently identify key input variables for expensive, black-box functions.
method Proposes novel active learning acquisition functions targeting derivative-based global sensitivity measures (DGSMs) under Gaussian process surrogate models.
result Active learning substantially enhances sample efficiency of DGSM estimation, especially with limited evaluation budgets.
In response to the development of recent efficient dense layers, this paper shows that something as simple as replacing linear components in pointwise convolutions with structured linear decompositions also produces substantial gains in the efficiency/accuracy tradeoff. Pointwise convolutions are fully connected layers…
Improved DOA estimation with distributed sensors across multiple frequencies.
problem Sensor gain uncertainties and directional perturbations in multi-frequency scenarios.
method Distributed optimization with local coherence models and iterative exchange of information.
result Advantages in statistical and computational efficiency through parallel iterative technique.
SCIENCE improves prediction intervals for individual causal effects.
problem Wide prediction intervals limit practical utility of causal inference.
method Surrogate-assisted conformal inference for efficient individual causal effects.
result SCIENCE produces more efficient prediction intervals for individual causal effects.
This paper improves parameter estimation in cardiac models using Gaussian process-based MH sampling.
problem Uncertainty in estimating patient-specific model parameters from sparse and noisy clinical data.
method Integrates surrogate modeling into Metropolis-Hastings sampling to improve computational efficiency and accuracy.
result Significant gain in computational efficiency without compromising accuracy, and insights into tissue heterogeneity.
Neural approximate computing gains enormous energy-efficiency at the cost of tolerable quality-loss. A neural approximator can map the input data to output while a classifier determines whether the input data are safe to approximate with quality guarantee. However, existing works cannot maximize the invocation of the a…
It is common to subsample Markov chain output to reduce the storage burden. Geyer (1992) shows that discarding k−1 out of every k observations will not improve statistical efficiency, as quantified through variance in a given computational budget. That observation is often taken to mean that thinning MCMC output ca…
DLM-One speeds up language generation by 500x with continuous models.
problem Efficiently generating text sequences in natural language processing.
method Score-distillation of continuous diffusion language models.
result Achieves up to 500x speedup in inference time with competitive performance.
Delayed rejection HMC improves sampling efficiency for multiscale distributions.
problem Hamiltonian Monte Carlo struggles with wide-ranging distributions, especially in high-curvature areas.
method Introduces a delayed rejection variant of HMC, using geometrically smaller step sizes for retries.
result Up to five-fold performance gains in effective sample size per gradient evaluation.
It is known that from purely observational data, a causal DAG is identifiable only up to its Markov equivalence class, and for many ground truth DAGs, the direction of a large portion of the edges will be remained unidentified. The golden standard for learning the causal DAG beyond Markov equivalence is to perform a se…
DLC enhances distillation-based continual learning with lightweight plugins.
problem Stability-plasticity dilemma in distillation-based continual learning.
method DLC deploys lightweight residual plugins into the classifier-proximal layer of a shared feature extractor.
result Significant 8% accuracy gain on large-scale benchmarks with minimal parameter increase.
Incremental versions of batch algorithms are often desired, for increased time efficiency in the streaming data setting, or increased memory efficiency in general. In this paper we present a novel algorithm for incremental kernel PCA, based on rank one updates to the eigendecomposition of the kernel matrix, which is mo…
We present Rotated Adaptive Tetra-iterated Quantizer (RATQ), a fixed-length quantizer for gradients in first order stochastic optimization. RATQ is easy to implement and involves only a Hadamard transform computation and adaptive uniform quantization with appropriately chosen dynamic ranges. For noisy gradients with al…
Novel neural architecture improves Bayesian experimental design efficiency.
problem Intractable evaluation of expected information gain (EIG) in Bayesian optimal experimental design.
method Develops a neural architecture that optimizes a single variational model for estimating EIG across many designs, using a lower bound for computational efficiency.
result Significantly improves accuracy in Bayesian experimental design with better sample efficiency.
We present a sparse grid high-order alternating direction implicit (ADI) scheme for option pricing in stochastic volatility models. The scheme is second-order in time and fourth-order in space. Numerical experiments confirm the computational efficiency gains achieved by the sparse grid combination technique.
A new causal graph framework identifies treatment effects without adjusting for confounders.
problem Invalid identification of causal effects due to unmeasured confounders.
method Developed the Napkin graph to identify causal effects through a ratio of g-formulas, using influence-function-based estimators.
result Demonstrated substantial efficiency gains in estimating causal effects using the Napkin graph.
New method uses diffusion models to optimize experimental design efficiently.
problem Optimizing experimental design for high-dimensional and complex settings.
method Introduces a pooled posterior distribution and uses diffusion-based samplers for efficient sampling and optimization.
result Extends Bayesian Optimal Experimental Design to practical scenarios.
Bayesian approach improves Shapley value estimation efficiency.
problem Efficiently estimating Shapley values in machine learning models.
method Bayesian experimental design using Gaussian process surrogate and adaptive coalition selection.
result Consistently improves sample efficiency in low-budget settings.
Gradient-free framework for Bayesian experimental design in complex systems.
problem Optimal experimental design in systems where gradient information is unavailable.
method Combines EKI and ALDI for optimization and sampling, with approximations for scalable utility estimation.
result Demonstrates robust, accurate, and efficient experimental design in various complex systems.
We approximate differential entropy for efficient Bayesian experimental design.
problem Efficiently estimating expected information gain in large-scale inference problems.
method Approximate differential entropy using Monte Carlo or quasi-Monte Carlo surrogates.
result Our approach achieves comparable or better convergence rates than state-of-the-art methods.
Ensembles of classification and regression trees remain popular machine learning methods because they define flexible non-parametric models that predict well and are computationally efficient both during training and testing. During induction of decision trees one aims to find predicates that are maximally informative …
iMOCA optimizes multiple objectives with continuous approximations for resource efficiency.
problem Optimizing multiple objectives with continuous function approximations that balance accuracy and evaluation cost.
method Information-Theoretic Multi-Objective Bayesian Optimization with Continuous Approximations (iMOCA) selects input and function approximations to maximize information gain per unit cost.
result iMOCA significantly improves over existing single-fidelity methods in approximating the optimal Pareto set.
Hypermodels improve exploration efficiency and accuracy.
problem Efficiently approximating Thompson sampling with large ensembles.
method Introducing hypermodels as a generalization of ensembles, including linear and neural network hypermodels.
result Hypermodels enable more accurate exploration and performance gains over Thompson sampling.
We propose randomized least-squares value iteration (RLSVI) -- a new reinforcement learning algorithm designed to explore and generalize efficiently via linearly parameterized value functions. We explain why versions of least-squares value iteration that use Boltzmann or epsilon-greedy exploration can be highly ineffic…
Paper proposes an unbiased optimization method for Bayesian experimental design.
problem Maximizing expected information gain in Bayesian experimental design.
method Randomized multilevel Monte Carlo (MLMC) method combined with stochastic gradient descent.
result An unbiased estimator for the gradient of expected information gain.
Introduces relative information gain for improving Gaussian process regression rates.
problem Improving the sample complexity of estimating or maximizing unknown functions.
method Introduces relative information gain, interpolates between effective dimension and information gain, and proves PAC-Bayesian bounds.
result Obtains minimax-optimal rates of convergence through the relative information gain.
We investigate the adversarial bandit problem with multiple plays under semi-bandit feedback. We introduce a highly efficient algorithm that asymptotically achieves the performance of the best switching m-arm strategy with minimax optimal regret bounds. To construct our algorithm, we introduce a new expert advice alg…
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.
Improves trial efficiency by adjusting for historical prognostic scores.
problem Reducing statistical uncertainty in randomized trial estimates.
method Linear covariate adjustment using a prognostic model trained on historical data.
result Prognostic covariate adjustment achieves minimum variance and reduces mean-squared error.
Decision trees algorithms use a gain function to select the best split during the tree's induction. This function is crucial to obtain trees with high predictive accuracy. Some gain functions can suffer from a bias when it compares splits of different arities. Quinlan proposed a gain ratio in C4.5's information gain fu…
Experimental design is crucial for inference where limitations in the data collection procedure are present due to cost or other restrictions. Optimal experimental designs determine parameters that in some appropriate sense make the data the most informative possible. In a Bayesian setting this is translated to updatin…