Relaxed Efficient Acquisition of Context and Temporal Features
Authors: Yunni Qu, Dzung Dinh, Grant King, Whitney Ringwald, Bing Cai Kok, Kathleen Gates, Aidan Wright, Junier Oliva
Organizations: The University of North Carolina at Chapel Hill · University of North Carolina at Chapel Hill, Chapel Hill, NC, USA · University of Michigan · University of Michigan, Ann Arbor, MI, USA · University of Minnisota Twin Cities · University of Minnesota Twin Cities, Minneapolis, MN, USA
In many biomedical applications, measurements are not freely available at inference time: each laboratory test, imaging modality, or assessment incurs financial cost, time burden, or patient risk. Longitudinal active feature acquisition (LAFA) seeks to optimize predictive performance under such constraints by adaptively selecting measurements over time, yet the problem remains inherently challenging due to temporally coupled decisions (missed early measurements cannot be revisited, and acquisition choices influence all downstream predictions). Moreover, real-world clinical workflows typically begin with an initial onboarding phase, during which relatively stable contextual descriptors (e.g., demographics or baseline characteristics) are collected once and subsequently condition longitudinal decision-making. Despite its practical importance, the efficient selection of onboarding context has not been studied jointly with temporally adaptive acquisition. We therefore propose REACT (Relaxed Efficient Acquisition of Context and Temporal features), an end-to-end differentiable framework that simultaneously optimizes (i) selection of onboarding contextual descriptors and (ii) adaptive feature--time acquisition plans for longitudinal measurements under cost constraints. REACT employs a Gumbel--Sigmoid relaxation with straight-through estimation to enable gradient-based optimization over discrete acquisition masks, allowing direct backpropagation from prediction loss and acquisition cost. Across real-world longitudinal health and behavioral datasets, REACT achieves improved predictive performance at lower acquisition costs compared to existing longitudinal acquisition baselines, demonstrating the benefit of modeling onboarding and temporally coupled acquisition within a unified optimization framework.
Figures & tables
Figure 1. Overview of REACT. REACT first performs a one-time onboarding acquisition of contextual descriptors at t=0 . Shaded feature entries denote acquired measurements. Using acquired context and prior longitudinal observations, the planner then adaptively decides both when to acquire next and which features to obtain. Between acquisition steps, the model predicts temporal labels without collecting additional measurements, yielding personalized, nonuniform acquisition trajectories.
Figure 2. Longitudinal planner. Given the onboarding context and longitudinal measurements observed up to time t , the planner outputs a binary acquisition mask over future feature–time pairs. The left portion of the mask corresponds to previously acquired measurements (unused), while the right portion specifies the future acquisition plan, judged against the cost-benefit objective. The earliest selected future time defines the next acquisition time, tnext .
Figure 3. AUROC/total cost of models across various average acquisition costs (budgets) on test data. The dashed line is evaluating the pretrained classifier from REACT with all features available.
Figure 4. Example feature acquisition rollouts using REACT for two distinct instances from the (a) CHEEARS and (b) WOMAC test sets. For visual clarity, only the selected features for these instances are displayed.
Figure 5. Qualitative visualization of longitudinal acquisition dynamics by REACT across datasets. Metrics are reported per dataset on the test set as (Total/Longitudinal Costs ∣ AUROC/AUPRC): ILIADD (35.739/23.993 ∣ 0.842/0.706), CHEEARS (13.275/11.275 ∣ 0.673/0.540), ADNI (12.635/10.535 ∣ 0.823/0.678), WOMAC (30.217/28.141 ∣ 0.670/0.355), KLG (29.243/26.869 ∣ 0.812/0.621). For visual clarity, only the selected features are displayed, and the stop probabilities are rounded.
Figure 6. Qualitative comparison of acquisition of REACT, DIME, and RAS on the ADNI dataset. Metrics are reported per framework on the test set as (Total/Longitudinal Costs ∣ AUROC/AUPRC): REACT (5.335/3.235 ∣ 0.824/0.683), DIME (5.619/3.630 ∣ 0.644/0.483), RAS (8.675/4.193 ∣ 0.817/0.684)
Figure 7. Ablation study of the self-iterative training we used for training REACT (Alg. 1 ). We compare AUROC/total cost when planner πθ trained with (i) self-iterative training from random initialization (Self Training) and (ii) trained using only offline reference states (Offline Reference).
Figure 8. Ablation study on how the learned context selector affects the temporal feature acquisition. AUROC of REACT with learned context descriptors compared to when context descriptors are all acquired (REACT-all) or not at all acquired (REACT-none) across datasets.
Appendix figures & tables15 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 9. AUPRC/total cost of models across various average acquisition costs (budgets) on test data. The dashed line is evaluating the pretrained classifier from REACT with all features available.
Figure 10. Ablation study of the self-iterative training for REACT (Alg. 1 ). We compare AUPRC/total cost when planner πθ trained with (i) self-iterative training from random initialization (Self Training) and (ii) trained using only offline reference states (Offline Reference).
Figure 11. Ablation study on how the learned context selector affects the temporal feature acquisition. AUPRC of REACT with learned context descriptors compared to when context descriptors are all acquired (REACT-all) or not at all acquired (REACT-none) across datasets.
Figure 12. Ablation on States Seen During Policy Training.
Dataset
λ
Total Cost
Temporal Cost
Context Cost
ADNI
0.05
1.100
0.500
0.600
0.01
4.725
3.225
1.500
0.0001
5.585
3.485
2.100
CHEEARS
0.003
1.000
0.000
1.000
0.002
6.851
5.851
1.000
0.0015
13.275
11.275
2.000
Appendix
Table 1. REACT hyperparameter: acquisition costs and λ by dataset.
Figure 13. Comparison of training and evaluation runtimes.
Feature Name
Acquisition Cost
Description
sex
1.0
Participant sex.
age
1.0
Participant age.
race___4
1.0
Participant race (Black/African American indicator).
hispanic
1.0
Hispanic/Latino ethnicity.
education
1.0
Highest level of education attained.
current_employed
1.0
Current employment status.
Appendix
Table 2. CHEEARS contextual descriptors.
Feature Name
Acquisition Cost
Description
happy
1.0
Self-reported happiness in the past 15 minutes (continuous scale).
nervous
1.0
Self-reported nervousness in the past 15 minutes (continuous scale).
angry
1.0
Self-reported anger in the past 15 minutes (continuous scale).
sad
1.0
Self-reported sadness in the past 15 minutes (continuous scale).
excited
1.0
Self-reported excitement in the past 15 minutes (continuous scale).
alert
1.0
Self-reported alertness in the past 15 minutes (continuous scale).
Event-contingent report of a social interaction (triggered by participant).
positiveaffEMA
1.0
Self-reported positive affect in the past 15 minutes (0 = Neutral, 10 = Very Positive).
energyEMA
1.0
Self-reported energy/alertness in the past 15 minutes (0 = Not at all, 10 = Very Energetic/Awake).
stressEMA
1.0
Self-reported stress in the past 15 minutes (0 = Not at all, 10 = Extremely).
impulse1EMA
1.0
In the past hour, acted on impulse (0 = Not at all, 10 = Very Much).
impulse2EMA
1.0
In the past hour, did things without worrying about consequences (0 = Not at all, 10 = Very Much).
Appendix
Table 5. ILIADD temporal (EMA) features.
Feature Name
Acquisition Cost
Description
AGE
0.3
Participant age.
PTGENDER
0.3
Participant gender.
PTEDUCAT
0.3
Participant education.
PTETHCAT
0.3
Participant ethnic category.
PTRACCAT
0.3
Participant racial category.
PTMARRY
0.3
Participant marital status.
Appendix
Table 6. ADNI contextual descriptors.
Feature Name
Acquisition Cost
Description
FDG
1.0
FDG PET biomarker.
AV45
1.0
AV45 PET biomarker.
Hippocampus
0.5
Hippocampal MRI biomarker.
Entorhinal
0.5
Entorhinal MRI biomarker.
Appendix
Table 7. ADNI temporal features.
Feature Name
Acquisition Cost
Description
HISP
0.3
Hispanic/Latino ethnicity indicator.
RACE
0.3
Self-reported race category.
SEX
0.3
Participant sex.
FAMHXKR
0.3
Mother, father, sister, or brother had knee repl surgery where all/part of knee replaced.
EDCV
0.3
Highest grade or year of school completed.
AGE
0.3
Age.
Appendix
Table 8. OAI contextual descriptors.
Feature Name
Acquisition Cost
Description
DRNKAMT
0.3
Current alcohol consumption amount.
DRKMORE
0.3
Alcohol-use indicator capturing heavier or more frequent drinking pattern.
BPSYS
0.5
Systolic blood pressure.
BPDIAS
0.5
Diastolic blood pressure.
BMI
0.5
Body mass index.
CEMPLOY
0.3
Current employment.
Appendix
Table 9. OAI temporal features.
Figure 14. Selected contexts for the policies shown in Fig. 5 . Bright cells indicate selected features, while dark cells indicate those that were not selected.
Accurate forecasting of pathological outcomes is a central problem in psychology. To do so, psychologists often collect intensive longitudinal data. However, in such studies, the desire to acquire a large number of variables for the sake of accurate prediction is often counteracted by the need to minimize participant burden. Acquiring more variables per occasion can yield better predictions, but having too many acquisitions increase the risk of non-response and attrition. Longitudinal Active Feature Acquisition (LAFA) is a principled approach to resolve this conundrum. Instead of requiring responses to every item at every acquisition occasion, LAFA produces a policy that seeks to optimally select dynamic subsets of items to be acquired at each timepoint while preserving our ability to forecast a specific outcome. However, existing LAFA methods are mostly based on Neural Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy. Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy.
Yunni Qu, Bing Cai Kok, Whitney Ringwald +4
Department of Computer Science, University of North Carolina at Chapel Hill · Department of Psychology and Neuroscience, University of North Carolina at Chapel Hill · School of Social Sciences, Nanyang Technological University, Singapore +2
Active Feature Acquisition (AFA) sequentially selects features to maximize predictive accuracy under an acquisition budget. A natural heuristic greedily maximizes Expected Information Gain minus a cost penalty, EIG(j|x_S)-lamc_j (EIG-Cost), but its theoretical guarantees are unclear. We provide an a-posteriori certificate decomposing the gap to the optimal hard-budget policy into a budget-slack term and a greedy-suboptimality term Del_lam. The certificate is computable when the penalized optimum is tractable; on small heterogeneous-cost instances, we verify it exactly, finding Del_lam near zero across 4,000 solved cases. On the large clinical task, Del_lam is not computable, so the decomposition is diagnostic rather than a numerical guarantee. Our analysis identifies EIG-Cost as one-step greedy optimization of the Lagrangian relaxation of expected-cost AFA, with lam as its dual multiplier. For equal costs, we prove Del_lam <= exp(-gamI(Y;X_[d])), where gam is the submodularity ratio, yielding a structural guarantee in the near-submodular regime. Supporting results establish an adaptive chain-rule identity showing that cumulative stepwise EIG optimizes transcript information, and prove that A -> I(Y;X_A) need not be submodular, even for jointly Gaussian variables. This rules out a direct Nemhauser-Wolsey-Fisher bound; naive Bayes restores submodularity. We also bound the randomization gap between hard-budget and expected-cost optima, analyze sensitivity to dual-multiplier error under local smoothness, and derive a conditional CVAE generator-shift bound given a bound on forward predictive KL. We treat this KL as a diagnostic quantity, not as a transform of training loss. Experiments cover synthetic Gaussian settings, AFA benchmarks, and clinical differential diagnosis on MIMIC-IV. On MIMIC-IV, EIG-Cost achieves macro-F1 0.180 +/- 0.012 at a $30 budget, using 6.6% of full-panel cost.
Joseph Bingham
Faculty of Computer Science Faculty of Biology Technion – Israel Institute of Technology Haifa 3200003, Israel
Active feature acquisition (AFA) asks which unobserved feature to measure next for each test instance under a budget. Greedy rules are easy to train but can overlook context features whose value is realized only through later acquisitions, while reinforcement-learning and generative approaches introduce difficult optimization or conditional-density estimation. We introduce \method, a deployable, supervised alternative that learns a separate candidate-conditioned risk-to-go function for every remaining budget. Starting from the one-step terminal classification risk, the functions are fitted backward with Bellman targets; inference greedily minimizes the learned terminal risk using only observed values, the mask, candidate identity, and remaining budget. A controlled non-myopic benchmark shows the expected mechanism: at budgets two and three, \method improves accuracy over its one-step ablation by 4.84±2.17 and 4.39±1.10 percentage points (mean ± standard error over five seeds). On Fashion-MNIST with 20 candidate pixels, it improves accuracy at every nontrivial reported budget on average, including 10.20±0.74 points at four acquisitions; its mean paired gain across budgets {2,4,8,12,16} is 3.50±0.37 points. A three-seed MiniBooNE study is mixed at small budgets but positive at 8 and 16 acquisitions, identifying a current boundary rather than supporting a universal claim. These results establish a reproducible mechanism-level case for direct Bellman risk regression and delimit the experiments still needed for state-of-the-art comparison.
Jiaorong Feng, Qian Li, Ying Li
Curtin Business School, Curtin University, Perth, Western Australia, Australia · †Present affiliation: Independent Researcher. · School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth, Western Australia, Australia