A framework uses a mixture of predictors for semi-supervised inference.
problem Limited labeled data, abundant unlabeled data.
method Mixture of Experts (MOE) for semi-supervised inference.
result MOE-powered inference framework achieves smallest possible variance.
Valid inference from data and predictions.
problem Valid statistical inference with machine learning predictions.
method Framework for valid inference using machine learning predictions.
result Valid confidence intervals without assumptions on predictions.
Paper extends prediction-powered inference using conformal prediction for robust and valid imputation.
problem Safe use of black-box ML models for imputing missing data with strong guarantees.
method Connecting prediction-powered inference with conformal prediction for valid and additional guarantees.
result First general prediction-powered procedure for e-values operating off-line.
Predictive e-values enhance statistical inference across various tasks.
problem Insufficient data limits traditional statistical inference.
method Apply prediction-powered inference to e-values.
result Every e-value-based inference has a prediction-powered counterpart.
FPPI selectively uses predictions to improve inference efficiency.
problem Improving statistical inference with limited labeled data and heterogeneous prediction quality.
method Filtered Prediction-Powered Inference (FPPI) framework.
result FPPI achieves strictly improved asymptotic efficiency compared to existing methods.
This work optimizes DNN inference for energy-harvesting devices by compressing and selectively executing neural network exits.
problem Inference delays and energy inefficiency in energy-harvesting devices.
method Developed a power trace-aware and exit-guided network compression algorithm for multi-exit neural networks.
result Superior accuracy and reduced latency compared to state-of-the-art techniques.
PPBoot simplifies prediction-powered inference.
problem Prediction-powered inference problems.
method Bootstrap-based method for arbitrary estimation problems.
result PPBoot often performs nearly identically to PPI(++).
We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate encoder outputs) and eliminating the redundant vectors. b) determining which word-vectors to eliminate …
PPI uses proxy data to improve inference from limited labels across related tasks.
problem Statistical inference with limited labels across multiple related tasks.
method Prediction-powered inference framework that uses cross-task recalibration to improve power and accuracy.
result Cross-task recalibration can substantially reduce confidence interval widths when labels are scarce.
PPI++ uses machine learning predictions to improve inference from small datasets.
problem Efficient inference from small labeled datasets with high-quality predictions.
method Adapts prediction-powered inference (PPI) to compute confidence sets for any parameter dimensionality.
result Improves classical intervals using only labeled data, always yielding better results.
Develops a more powerful selective inference method for stepwise feature selection.
problem Loss of power in existing conditional SI methods due to over-conditioning.
method Uses homotopy continuation approach to overcome over-conditioning.
result Shows improved power and efficiency in selective inference for feature selection.
Cross-prediction improves inference from small labeled datasets.
problem Valid inference from small labeled datasets with imperfect predictions.
method Imputes missing labels via machine learning and debiases predictions.
result Inferences achieve desired error probability and are more powerful.
FAB-PPI uses prior knowledge to improve prediction-powered inference.
problem Improving statistical inference with machine learning predictions.
method Informing PPI with prior knowledge on prediction quality.
result FAB-PPI improves inference accuracy and confidence intervals.
We present a model for random simple graphs with a degree distribution that obeys a power law (i.e., is heavy-tailed). To attain this behavior, the edge probabilities in the graph are constructed from Bertoin-Fujita-Roynette-Yor (BFRY) random variables, which have been recently utilized in Bayesian statistics for the c…
Extends PPI to sequential setting, improving inference over time.
problem Sequential data growth with unlabelled data.
method Prediction-powered confidence sequence procedures using Ville's inequality and the method of mixtures.
result Asymptotically valid uniformly over time, accommodating prior knowledge.
Exact selective inference with randomization for Gaussian regression models.
problem Exact selective inference in Gaussian regression models.
method Introduces a pivot for exact selective inference with randomization, reducing the problem to a bivariate truncated Gaussian distribution.
result Our pivot leads to exact inference and produces narrower confidence intervals than related methods.
PPI uses predictions and weighting to infer from partially labeled data.
problem Valid inference with partially labeled data.
method Combines model-based predictions with bias correction from labeled data, using Horvitz-Thompson and Hájek corrections.
result IPW-adjusted PPI with estimated propensities performs similarly to known-probability case.
Improved local multivariable regression for better inference with limited data.
problem Limited sample size hampers local polynomial/multivariable regression.
method Prediction-Powered Inference (PPI) algorithm for local multivariable regression.
result Significantly reduces estimation variance without increasing error.
Improved Bayesian inference using power priors with historical data.
problem Improving Bayesian inference with historical data.
method Generalized power priors that adapt to the α parameter of Amari's α-divergence. result Improved performance through appropriate choices of the α parameter. Imputation-Powered Inference improves subpopulation efficiency in missing data settings.
problem Complex missing data patterns challenge standard inference methods.
method Imputation-Powered Inference (IPI) combines blackbox imputation with bias correction.
result IPI provides valid and efficient M-estimation under MCAR blockwise missingness.
StratPPI improves prediction-powered inference with stratified sampling.
problem Improving statistical estimates with limited human-labeled data.
method Combining small human-labeled data with large automatic-labeled data, stratifying data for tighter confidence intervals.
result StratPPI provides substantially tighter confidence intervals than unstratified approaches.
The ever-increasing demand from mobile Machine Learning (ML) applications calls for evermore powerful on-chip computing resources. Mobile devices are empowered with heterogeneous multi-processor Systems-on-Chips (SoCs) to process ML workloads such as Convolutional Neural Network (CNN) inference. Mobile SoCs house sever…
Empirical evidence suggests that heavy-tailed degree distributions occurring in many real networks are well-approximated by power laws with exponents η that may take values either less than and greater than two. Models based on various forms of exchangeability are able to capture power laws with η<2, and admit tra…
Enhances selective inference for generalized lasso using parametric programming.
problem Low statistical power in selective inference for generalized lasso.
method Parametric programming to compute solution paths and identify model selection events.
result Improves selective inference power and practicality for various problems.
Enhances statistical inference using synthetic data.
problem Limited labeled data for statistical inference.
method GESPI framework that combines synthetic and real data.
result Error rate remains below a user-specified bound and decreases with synthetic data quality.
Identifying a potentially large number of simultaneous line outages in power transmission networks in real time is a computationally hard problem. This is because the number of hypotheses grows exponentially with the network size. A new "Learning-to-Infer" method is developed for efficient inference of every line statu…
Using Intel's Loihi neuromorphic research chip and ABR's Nengo Deep Learning toolkit, we analyze the inference speed, dynamic power consumption, and energy cost per inference of a two-layer neural network keyword spotter trained to recognize a single phrase. We perform comparative analyses of this keyword spotter runni…
AM-PPI uses multiple predictors to reduce label cost in healthcare AI.
problem Reduces label cost in post-deployment monitoring of healthcare AI.
method Combines model predictions with a small labeled sample, routing each instance to a cost-appropriate subset of predictors.
result Produces narrower confidence intervals than single-predictor methods.
Enhances weak lensing inference with neural summaries.
problem Extracting additional information from weak lensing convergence maps.
method Hybrid approach combining physics-based and neural summaries.
result Neural summaries extract up to 8 times more information than angular power spectra.
Variational inference is a powerful tool for approximate inference, and it has been recently applied for representation learning with deep generative models. We develop the variational Gaussian process (VGP), a Bayesian nonparametric variational family, which adapts its shape to match complex posterior distributions. T…
Study improves treatment effect estimation using unlabeled covariates.
problem Estimating treatment effects with limited labeled data.
method Developed efficiency bounds and estimators for semi-supervised setting.
result Estimators using unlabeled covariates have lower asymptotic variance.
New Stein operator improves robustness in model inference.
problem Improving robustness in inference for unnormalized models.
method Density-power weighted Stein operator (γ-Stein operator). result Robust methods for goodness-of-fit testing and posterior approximation.
Enhances multi-modular models by directing information flow between components.
problem Improving predictive performance in multi-modular models with misspecification.
method Introduces Semi-Modular Inference (SMI) with an influence parameter to control information flow between modules.
result SMI allows for tunable and directed information flow, improving prediction in some settings.
We study two aspects of noisy computations during inference. The first aspect is how to mitigate their side effects for naturally trained deep learning systems. One of the motivations for looking into this problem is to reduce the high power cost of conventional computing of neural networks through the use of analog ne…
Generalizes prediction-powered inference for binary classifier evaluation.
problem Evaluation of binary classifiers with partially observed outcomes.
method Generalizes PPI to any regular asymptotically linear estimator and proposes modified estimators for covariate shift.
result PPI can be a computationally-simple alternative to existing methods, achieving no greater than the semi-parametric efficiency lower bound in certain scenarios.
Paper establishes statistical inference for performative predictions.
problem Dynamic influence of predictions on their targets.
method End-to-end framework for estimation and inference under performativity.
result Established central limit theorem for performative settings.
We explore the robustness of recurrent neural networks when the computations within the network are noisy. One of the motivations for looking into this problem is to reduce the high power cost of conventional computing of neural network operations through the use of analog neuromorphic circuits. Traditional GPU/CPU-cen…
New method detects if data points were used in training models with low cost and high power.
problem Detecting if a particular data point was used in training a model.
method Fine-grained modeling of null hypothesis in likelihood ratio tests, leveraging reference models and population data.
result RMIA has superior test power compared to prior methods, even at extremely low false positive rates.
A new method for kernel tests without data splitting increases power.
problem Lack of power in kernel-based tests due to data splitting.
method Selective inference framework to learn hyperparameters and test on full sample.
result Empirically larger test power without data splitting, regardless of split proportion.
PPI uses predictions to improve inference from incomplete data.
problem Incomplete or costly-to-measure outcomes in research fields.
method Leverages large unlabeled datasets for improved statistical efficiency with bias correction.
result PPI variants produce tighter confidence intervals than complete-case analysis.
Active inference uses machine learning to prioritize data labeling for more efficient statistical inference.
problem Efficiently collecting data points for statistical inference with limited labels.
method A machine learning-assisted approach that identifies uncertain data points for labeling.
result Achieves the same level of accuracy with fewer samples, resulting in smaller confidence intervals and more powerful p-values.
Novel conformal methods test out-of-distribution data with labeled outliers.
problem Testing whether new data comes from the same distribution as a reference.
method Integrative conformal p-values re-weight standard p-values using dependent side information.
result The methods outperform standard conformal p-values in simulations and applications.
New method uses machine learning to improve statistical inference.
problem Performing inference on conditional functionals with scarce labeled data.
method Combines localization with prediction-based variance reduction.
result Valid and sharp confidence intervals for conditional functionals.
PAS improves estimation of multiple means using ML predictions and shrinkage.
problem Improving statistical estimates with limited gold-standard data and noisy ML predictions.
method Prediction-Powered Adaptive Shrinkage (PAS) that combines PPI with empirical Bayes shrinkage.
result PAS adapts to the reliability of ML predictions and outperforms traditional methods in large-scale applications.
Book introduces ML and AI for causal inference.
problem Uncertainty in causal relationships.
method Structural equation models, DAGs, SCMs, and Double/Debiased Machine Learning.
result Improved inference in causal models using predictive tools.
OptCS optimizes model selection after conformal inference, controlling FDR and power loss.
problem Challenges in model selection for conformal inference, especially when limited labeled data and many model choices are available.
method OptCS framework that allows valid statistical testing after flexible data-driven model optimization, using novel multiple testing procedures.
result Valid conformal p-values constructed despite substantial data reuse, maintaining FDR control.
New method finds significant high-order interactions efficiently.
problem Finding statistically significant high-order interactions in high-dimensional data.
method Extends selective inference to high-order interaction models with pruning strategy.
result Demonstrated efficient and powerful method for high-order interactions.
A power-law fit to the empirical inference-compute frontier in LOB prediction suggests a scaling-law-style frontier.
problem Limit order book prediction
method Using a suite of models ranging from small decision trees to neural LOB architectures
result A power-law fit to the low- and mid-compute non-MLPLOB frontier extrapolates across multiple orders of magnitude and attains R2=0.941 on the excluded high-compute MLPLOB target frontier.