Meta tags:
Headings (most frequently used words):
causal, ml, book, research, group, publications, and, working, papers, nathan, kallus, nah, tahn, kah, loosh,
Text of the page (most frequently used words):
the (1413), and (1105), for (412), with (384), that (368), this (247), policy (236), data (227), learning (224), inference (204), our (203), #causal (191), optimization (176), abstract (167), arxiv (155), decision (152), are (150), from (140), can (126), treatment (122), which (119), personalization (119), under (119), based (117), model (107), optimal (104), when (103), study (101), making (100), sequential (97), using (96), show (95), conference (92), these (90), approach (88), problem (88), such (81), new (80), methods (80), propose (79), where (79), not (76), bounds (76), evaluation (75), function (74), effects (73), efficient (70), estimation (69), both (68), method (67), reinforcement (67), may (65), models (64), estimators (63), provide (62), term (62), machine (60), effect (59), algorithm (58), policies (58), value (57), regret (57), paper (56), robust (56), estimator (54), demonstrate (52), time (50), but (50), even (50), setting (50), off (50), 2025 (50), only (49), experiments (49), outcomes (49), conditional (49), their (48), while (48), best (48), systems (48), standard (47), observational (47), guarantees (47), class (47), framework (46), case (46), large (45), novel (45), algorithms (45), functions (45), long (45), over (44), many (44), than (44), error (44), asymptotic (44), each (43), one (43), sample (43), linear (43), performance (43), has (42), information (42), how (42), international (41), results (41), average (41), develop (41), existing (41), github (41), non (40), all (40), first (40), rates (40), 2024 (40), fairness (39), settings (39), problems (39), given (39), bound (39), estimate (38), its (38), two (38), confounding (38), outcome (37), work (37), 2020 (37), minimax (36), neural (36), risk (36), ope (36), 2023 (36), offline (36), applied (35), state (35), use (35), prove (35), valid (35), conversational (35), often (34), more (33), personalized (33), approaches (33), efficiency (33), training (33), confounders (33), however (32), have (32), observations (32), conditions (32), reward (32), neurips (32), between (32), loss (32), processing (31), analysis (31), well (31), management (30), predictive (30), recommendation (30), variance (30), contextual (30), regression (30), across (30), confidence (30), assumptions (29), also (29), any (29), 2022 (29), control (29), estimating (29), nonparametric (29), consider (28), high (28), february (28), research (28), 2018 (28), rate (28), bandit (28), inverse (28), adaptive (28), empirical (27), weighted (27), 2021 (27), unobserved (27), bellman (27), applications (26), number (26), design (26), observed (26), real (26), statistical (26), icml (26), distributional (26), via (26), into (25), actions (25), parameters (25), target (25), possible (24), learn (24), empirically (24), without (24), synthetic (24), variable (24), general (24), convergence (24), bias (24), different (24), treatments (24), 2019 (24), practice (23), coverage (23), other (23), sampling (23), identification (23), weighting (23), dependent (23), due (23), markov (23), single (22), individual (22), test (22), finite (22), asymptotically (22), distribution (22), order (22), journal (22), further (22), doubly (22), exploration (22), instrumental (22), nuisance (22), bibaut (22), language (21), theoretical (21), present (21), improve (21), mean (21), kernel (21), uncertainty (21), arm (21), estimates (21), process (21), bandits (21), continuous (21), multiple (21), semiparametric (21), uehara (21), they (20), behavior (20), low (20), structure (20), parametric (20), structural (20), matching (20), june (20), statistics (20), approximation (20), cost (20), some (20), weights (20), level (20), establish (20), condition (20), weak (20), fast (20), operations (19), practical (19), most (19), datasets (19), set (19), used (19), proposed (19), dynamic (19), latent (19), achieves (19), dataset (19), recent (19), sharp (19), surrogate (19), estimated (19), benefits (19), short (19), llms (19), including (18), world (18), experimental (18), lower (18), driven (18), functional (18), tests (18), strong (18), science (18), decisions (18), importance (18), classification (18), mao (18), introduce (18), experiment (18), horizon (18), experimentation (18), recommender (18), wang (18), llm (18), complexity (17), means (17), specific (17), dimensional (17), small (17), acm (17), scale (17), space (17), consistent (17), hypothesis (17), available (17), interest (17), probability (17), challenge (17), multi (17), rank (17), crucial (17), simulation (17), instead (17), action (17), zhou (17), complex (17), processes (17), feedback (17), interference (17), impact (16), future (16), power (16), assignment (16), form (16), particular (16), stochastic (16), baseline (16), prediction (16), balance (16), covariates (16), metrics (16), learned (16), extend (16), been (16), soft (16), point (16), group (16), mdps (16), stationary (16), debiased (16), sun (16), functionals (16), 2026 (16), fail (15), second (15), whether (15), improvement (15), recommendations (15), either (15), characterize (15), leverage (15), online (15), since (15), december (15), october (15), obtain (15), propensity (15), representation (15), directly (15), assumption (15), achieve (15), would (15), binary (15), leveraging (15), fixed (15), derive (15), intervals (15), observable (15), being (14), objective (14), randomization (14), focus (14), item (14), result (14), gradient (14), aistats (14), flexible (14), various (14), unknown (14), evaluate (14), rather (14), theory (14), double (14), user (14), task (14), does (14), sensitive (14), partially (14), address (14), appear (14), cate (14), requires (14), bennett (14), better (14), parameter (14), fitted (14), reasoning (14), simple (13), web (13), application (13), groups (13), randomized (13), intelligence (13), regularized (13), potential (13), solutions (13), testing (13), there (13), efficiently (13), validate (13), medicine (13), lead (13), interventions (13), instrument (13), safety (13), entropy (13), smooth (13), difference (13), then (13), rewards (13), nuisances (13), liang (13), completeness (13), after (13), construct (13), end (13), normality (13), knowledge (13), preferences (12), against (12), context (12), previous (12), through (12), natural (12), random (12), make (12), beyond (12), important (12), population (12), assortment (12), threshold (12), improvements (12), solution (12), historical (12), covariate (12), literature (12), evaluating (12), resulting (12), same (12), example (12), exposure (12), challenging (12), providing (12), enables (12), still (12), moment (12), safe (12), infinite (12), defined (12), realizability (12), mdp (12), normal (12), environment (12), pomdps (12), crs (12), collaborative (12), significantly (11), social (11), typically (11), controlled (11), discrete (11), choice (11), likelihood (11), care (11), popular (11), exact (11), artificial (11), computationally (11), properties (11), key (11), examples (11), distributions (11), generally (11), challenges (11), variety (11), identified (11), certain (11), misspecification (11), consistency (11), need (11), squared (11), require (11), correct (11), studies (11), november (11), limited (11), surrogates (11), remains (11), conduct (11), finally (11), cases (11), step (11), transition (11), factor (11), agnostic (11), van (11), der (11), laan (11), least (11), anytime (11), revenue (10), public (10), type (10), 2016 (10), observation (10), norm (10), maximum (10), patient (10), guarantee (10), august (10), measures (10), specifically (10), pricing (10), features (10), limitations (10), called (10), special (10), always (10), worst (10), restrictions (10), generalized (10), specified (10), discuss (10), shift (10), domain (10), generalization (10), compared (10), varying (10), presence (10), balancing (10), simultaneously (10), controlling (10), negative (10), states (10), true (10), spaces (10), classes (10), particularly (10), variables (10), disparity (10), uses (10), covariance (10), matrix (10), plug (10), sensitivity (10), adversarial (10), network (10), find (10), partial (10), very (10), within (10), way (10), cumulative (10), three (10), stage (10), fqi (10), zero (10), oprescu (10), pac (10), token (10), allocation (9), whose (9), cross (9), out (9), common (9), preference (9), clinical (9), diabetes (9), alternative (9), algorithmic (9), minimizing (9), tractable (9), saa (9), similar (9), provides (9), direct (9), units (9), must (9), necessary (9), limits (9), designs (9), items (9), good (9), free (9), robustness (9), representations (9), scores (9), apply (9), relevant (9), investigate (9), oracle (9), affected (9), biased (9), predictions (9), proxy (9), known (9), support (9), extensive (9), difficult (9), bail (9), allows (9), networks (9), gap (9), kallus (9), healthcare (9), moments (9), stability (9), because (9), terms (9), learns (9), near (9), upper (9), equation (9), splitting (9), deterministic (9), minimum (9), iteration (9), agent (9), bridge (9), march (9), unmeasured (9), identifying (9), tasks (9), change (9), search (9), implicit (9), steck (9), source (9), cho (9), several (8), extended (8), identify (8), adapt (8), bertsimas (8), content (8), media (8), event (8), role (8), selection (8), accuracy (8), predicting (8), mining (8), informs (8), samples (8), about (8), collected (8), include (8), past (8), allowing (8), customer (8), much (8), current (8), improves (8), health (8), similarity (8), motivated (8), sets (8), january (8), mild (8), enabling (8), coefficient (8), metric (8), nonetheless (8), precision (8), strategy (8), arms (8), assume (8), size (8), compare (8), regimes (8), marginal (8), handle (8), fully (8), them (8), rules (8), system (8), noise (8), generated (8), learner (8), hidden (8), yields (8), deep (8), estimands (8), provably (8), downstream (8), just (8), take (8), objectives (8), approximate (8), adaptively (8), post (8), product (8), embeddings (8), queries (8), index (8), netflix (8), influence (8), temporal (8), lambda (8), irl (8), main (7), moreover (7), significant (7), 2014 (7), widely (7), insights (7), range (7), sequence (7), www (7), trials (7), statistically (7), equivalent (7), smaller (7), september (7), 2017 (7), hilbert (7), outperform (7), rich (7), techniques (7), enjoys (7), upon (7), experience (7), modern (7), unique (7), enable (7), cannot (7), dependence (7), society (7), jrss (7), contexts (7), improved (7), reduction (7), split (7), reliable (7), pre (7), effectiveness (7), overcome (7), kom (7), violations (7), classic (7), armed (7), analyze (7), tackle (7), residual (7), fact (7), showing (7), art (7), property (7), assess (7), protected (7), labels (7), reduce (7), factorization (7), induced (7), like (7), percnt (7), modeling (7), leads (7), proxies (7), positive (7), instruments (7), components (7), normalized (7), performing (7), regularization (7), controls (7), targeted (7), contrast (7), highlight (7), exist (7), ratios (7), global (7), prior (7), therefore (7), recently (7), tabular (7), lee (7), constructing (7), understanding (7), changes (7), ate (7), cvar (7), debiasing (7), martingale (7), digital (7), classical (7), goal (7), cosine (7), log (7), calibration (7), compliance (7), unet (7), professor (7), business (6), operational (6), unified (6), formulation (6), big (6), open (6), what (6), targets (6), analytics (6), wide (6), too (6), magnitude (6), computational (6), code (6), involving (6), heterogeneous (6), underlying (6), relative (6), medical (6), maximize (6), operator (6), furthermore (6), illustrate (6), distributionally (6), outperforms (6), independent (6), nature (6), confounded (6), complete (6), developed (6), royal (6), series (6), central (6), advertising (6), types (6), connections (6), give (6), characterization (6), drug (6), instance (6), achieving (6), having (6), comparing (6), iptw (6), santacatterina (6), relies (6), human (6), numerical (6), focuses (6), ipw (6), leverages (6), benchmark (6), benchmarks (6), thus (6), assessing (6), simulations (6), noisy (6), values (6), procedure (6), optimize (6), overlap (6), latter (6), inferences (6), effectively (6), estimand (6), gate (6), uniform (6), minimization (6), suggest (6), april (6), benefit (6), disparate (6), impacts (6), response (6), account (6), ranking (6), define (6), involve (6), bounded (6), compute (6), retargeting (6), jasa (6), discussion (6), pooling (6), constrained (6), constraints (6), density (6), drl (6), feasible (6), trajectory (6), differentiable (6), quantile (6), meta (6), ratio (6), access (6), gains (6), quality (6), xie (6), faster (6), depends (6), train (6), primary (6), trajectories (6), generic (6), reform (6), crime (6), regressions (6), sqrt (6), face (6), fqe (6), bayesian (6), mcinerney (6), users (6), rely (6), sequences (6), explicit (6), times (6), shot (6), fine (6), spatial (6), neighborhood (6), spatiotemporal (6), kdd (6), dvds (6), misspecified (6), stopping (6), reduces (6), baselines (6), automatic (6), agents (6), gst (6), assistant (6), university (6), scheduling (5), success (5), solving (5), crowd (5), before (5), amount (5), half (5), competition (5), analyzing (5), events (5), inventory (5), observe (5), dimension (5), economics (5), records (5), were (5), characteristics (5), history (5), predicted (5), expected (5), rise (5), arise (5), programming (5), perspective (5), returns (5), joint (5), efficacy (5), price (5), final (5), formalize (5), optimality (5), powerful (5), notion (5), previously (5), offers (5), effective (5), memory (5), related (5), gom (5), combining (5), gaussian (5), jmlr (5), tools (5), job (5), building (5), arises (5), above (5), score (5), kow (5), fitting (5), optimally (5), shown (5), conclusions (5), mab (5), adjust (5), accounting (5), question (5), unconfoundedness (5), exactly (5), tightest (5), missing (5), reason (5), facct (5), posterior (5), intervention (5), trained (5), generating (5), unlike (5), reference (5), made (5), regardless (5), explore (5), analyses (5), minimize (5), rule (5), review (5), avoids (5), yet (5), education (5), another (5), infeasible (5), snis (5), ensure (5), local (5), satisfy (5), affect (5), issue (5), discovery (5), involves (5), intuition (5), solve (5), possibly (5), fundamental (5), smoothness (5), combination (5), could (5), colt (5), depend (5), gradients (5), avoiding (5), oral (5), july (5), collection (5), welfare (5), goods (5), errors (5), environments (5), expectation (5), closedness (5), converge (5), pomdp (5), proximal (5), semiparametrically (5), credible (5), findings (5), naturally (5), quantification (5), scenarios (5), pairs (5), active (5), equations (5), feature (5), jia (5), jive (5), dynamics (5), mcauley (5), unit (5), optimism (5), graphs (5), mapping (5), convolution (5), jackknife (5), ivs (5), npiv (5), remain (5), distrl (5), limit (5), zenati (5), bilevel (5), grpo (5), link (5), academic (4), division (4), desirable (4), notably (4), fair (4), protests (4), cyber (4), sources (4), seen (4), winner (4), variances (4), orders (4), evidence (4), created (4), those (4), location (4), thousands (4), signals (4), traditional (4), ground (4), choices (4), planning (4), shows (4), specialized (4), computation (4), udell (4), factors (4), electronic (4), evaluated (4), dual (4), expressed (4), gives (4), quadratic (4), specification (4), availability (4), theoretically (4), enjoy (4), probabilistic (4), practitioners (4), whenever (4), finalist (4), 2013 (4), mathematical (4), fit (4), procedures (4), insight (4), numerically (4), developing (4), quantities (4), demand (4), auxiliary (4), measure (4), per (4), measured (4), posed (4), quantity (4), applying (4), priori (4), hand (4), hoc (4), block (4), demonstrated (4), formulate (4), suggests (4), scales (4), impose (4), surprisingly (4), few (4), interactions (4), converges (4), ones (4), applies (4), commonly (4), augmented (4), choose (4), consists (4), partition (4), generalize (4), sometimes (4), adaptation (4), shalit (4), extreme (4), stabilized (4), convex (4), informative (4), censoring (4), studying (4), positivity (4), spine (4), violated (4), issues (4), matches (4), argue (4), studied (4), unfairness (4), criteria (4), realistic (4), derived (4), respect (4), membership (4), disparities (4), why (4), strongly (4), interplay (4), accountability (4), transparency (4), principled (4), exponential (4), points (4), makes (4), finding (4), increasingly (4), usually (4), spotlight (4), candidate (4), around (4), nominal (4), worse (4), harm (4), orthogonality (4), koow (4), penalized (4), directions (4), costly (4), unbiased (4), optimizing (4), domains (4), never (4), variants (4), variational (4), higher (4), pure (4), needing (4), self (4), increase (4), although (4), american (4), association (4), independently (4), phenomenon (4), grows (4), importantly (4), unidentifiable (4), assumes (4), curse (4), consistently (4), minimal (4), carefully (4), initial (4), lax (4), reductions (4), yield (4), draw (4), implies (4), taking (4), steps (4), direction (4), projected (4), combined (4), explicitly (4), circumvent (4), logging (4), dtr (4), utility (4), zhao (4), sufficient (4), offer (4), generate (4), thought (4), potentially (4), accommodate (4), track (4), critic (4), might (4), here (4), allow (4), replacing (4), publicly (4), drope (4), vanilla (4), divergence (4), nonparametrically (4), measuring (4), inputs (4), during (4), computer (4), differences (4), focusing (4), becomes (4), works (4), negatively (4), tighter (4), projection (4), sabre (4), inequalities (4), regime (4), admit (4), avoid (4), variation (4), query (4), capabilities (4), traces (4), squares (4), guide (4), exhibit (4), syrgkanis (4), iclr (4), switchback (4), mse (4), aim (4), closed (4), indices (4), titles (4), tokens (4), recsys (4), filtering (4), experimenters (4), assuming (4), enhance (4), superior (4), signal (4), desired (4), strategies (4), cates (4), traffic (4), together (4), persistent (4), crave (4), adaptivity (4), scaling (4), explain (4), monitoring (4), a_t (4), b_t (4), hessian (4), autodml (4), orthogonal (4), valued (4), geometric (4), num (4), modular (4), mappings (4), directional (4), diffusion (4), dllms (4), demonstrations (4), eat (4), ddc (4), cgdp (4), bai (4), cornell (4), today (3), hospital (3), growing (3), burdensome (3), assignments (3), extent (3), simply (3), integer (3), radiology (3), flexibility (3), tractability (3), accessible (3), recorded (3), massive (3), extracted (3), occurrence (3), predict (3), 2015 (3), aims (3), discrepancy (3), extract (3), see (3), thereby (3), crucially (3), ubiquitous (3), offerings (3), takes (3), follow (3), separate (3), combinations (3), descent (3), chosen (3), matched (3), among (3), visits (3), substantial (3), embedding (3), depending (3), utilizing (3), concrete (3), gupta (3), hold (3), additionally (3), award (3), costs (3), associated (3), uncertain (3), equal (3), critical (3), imbalance (3), relying (3), ensures (3), consumer (3), commerce (3), larger (3), globally (3), aware (3), analytical (3), automatically (3), trade (3), partitioning (3), finds (3), workshop (3), incorporating (3), johansson (3), minimizes (3), decomposition (3), people (3), stable (3), able (3), surgical (3), pennicooke (3), iab (3), will (3), unrestricted (3), kind (3), captures (3), batched (3), rejection (3), almost (3), adjusting (3), actually (3), concern (3), induce (3), reweighting (3), stop (3), interval (3), vary (3), race (3), imputation (3), family (3), look (3), leading (3), tested (3), weaker (3), status (3), supports (3), happened (3), monotonicity (3), nets (3), subpopulation (3), easily (3), comparative (3), questions (3), consumption (3), jiang (3), dtrs (3), extends (3), program (3), tool (3), full (3), tuned (3), characterized (3), bipartite (3), xauc (3), fields (3), distance (3), necessity (3), individualized (3), changing (3), optimized (3), counseling (3), easy (3), frameworks (3), includes (3), far (3), reaching (3), implications (3), fold (3), invariant (3), ergodic (3), characterizes (3), variant (3), sufficiently (3), slow (3), achieved (3), running (3), sense (3), qte (3), whole (3), localized (3), favorable (3), subject (3), extending (3), help (3), major (3), targeting (3), stratified (3), line (3), equity (3), considerations (3), concerns (3), comprehensive (3), behavioral (3), budget (3), null (3), inequality (3), repeated (3), arrival (3), forest (3), every (3), construction (3), forests (3), shelf (3), arbitrarily (3), decays (3), supervised (3), participants (3), longer (3), poses (3), cadr (3), technical (3), uniquely (3), york (3), checks (3), less (3), solved (3), ite (3), heterogeneity (3), black (3), box (3), polynomial (3), chang (3), delta (3), wald (3), permits (3), calibrated (3), introduced (3), style (3), click (3), exposed (3), fraction (3), conservative (3), methodology (3), tail (3), pseudo (3), onto (3), comes (3), provable (3), appropriate (3), return (3), rejecting (3), concentration (3), seek (3), usprt (3), verify (3), lindon (3), setup (3), dorn (3), quasi (3), total (3), continuously (3), frazier (3), sak (3), attains (3), constant (3), joingym (3), quickly (3), 200 (3), move (3), surrogacy (3), ejdemyr (3), margin (3), pessimism (3), conversations (3), jha (3), following (3), obtaining (3), imposing (3), newey (3), concentrability (3), wise (3), satisfies (3), platforms (3), certainty (3), truncated (3), thompson (3), especially (3), mismatch (3), vectors (3), quantify (3), arbitrary (3), similarities (3), inspired (3), remove (3), recommended (3), popularity (3), should (3), answers (3), reddit (3), despite (3), comparable (3), delayed (3), technology (3), engagement (3), highly (3), computed (3), sigkdd (3), chou (3), tran (3), composite (3), streams (3), peak (3), gan (3), severely (3), conformal (3), gao (3), crag (3), zhu (3), schemes (3), sizes (3), cspi (3), detection (3), stringent (3), requirements (3), guardrail (3), simultaneous (3), augmentation (3), guided (3), clo (3), ierm (3), induces (3), losses (3), requirement (3), verbalized (3), inverting (3), maximization (3), reducing (3), alpha (3), requiring (3), updates (3), adjustment (3), firms (3), iterative (3), amriv (3), iii (3), alignment (3), math (3), maintaining (3), grained (3), vgs (3), majority (3), required (3), softmax (3), contrasts (3), aligning (3), scalar (3), gretton (3), furious (3), coupled (3), early (3), belief (3), preserving (3), cmf (3), filter (3), school (3), book (3), physician (2), expanding (2), schedule (2), naïve (2), unfair (2), adverse (2), consequences (2), incorporates (2), noting (2), disciplines (2), mixed (2), implementing (2), women (2), directed (2), increasing (2), becoming (2), shared (2), greater (2), activism (2), turning (2), continually (2), 300 (2), news (2), government (2), publications (2), occur (2), entities (2), involved (2), sentiment (2), tone (2), discussed (2), trends (2), clearly (2), exponentially (2), bank (2), development (2), johnson (2), designing (2), harvested (2), acts (2), timeframe (2), reported (2), situation (2), inferred (2), 2006 (2), production (2), products (2), ads (2), consisting (2), mixture (2), segment (2), revenues (2), revealed (2), guidelines (2), managing (2), patients (2), center (2), nearest (2), neighbor (2), switching (2), exceeds (2), hba1c (2), mmol (2), mol (2), conclusion (2), visualizing (2), inform (2), 217 (2), priors (2), embeds (2), reproducing (2), universal (2), kernels (2), last (2), part (2), paradigm (2), applicable (2), portfolio (2), conventional (2), 167 (2), modification (2), linking (2), connection (2), combine (2), ideas (2), prescribe (2), departure (2), reflecting (2), imperfect (2), distributed (2), termed (2), prescriptiveness (2), prescriptive (2), imdb (2), google (2), collect (2), leveraged (2), pose (2), assessment (2), purposes (2), validity (2), rigorous (2), ijoo (2), restricting (2), parametrically (2), recovers (2), rerandomization (2), nearly (2), perfect (2), biasing (2), advantages (2), retail (2), dynamically (2), static (2), learnable (2), behind (2), generalizing (2), interpretation (2), interpretability (2), monotonic (2), weight (2), options (2), interpretable (2), unsupervised (2), overcoming (2), unrealistic (2), knowing (2), devise (2), sontag (2), truncation (2), cbps (2), msm (2), addressing (2), living (2), hiv (2), sate (2), problematic (2), uniformly (2), spuriously (2), differ (2), regrets (2), alt (2), fails (2), reject (2), optimizer (2), dosing (2), connect (2), adjusted (2), systematic (2), classifiers (2), perpetuates (2), injustices (2), assessments (2), focused (2), hormone (2), replacement (2), therapy (2), trial (2), complicated (2), reasons (2), geolocation (2), explains (2), occurs (2), hard (2), mortgage (2), confirm (2), simpler (2), measurements (2), measurement (2), infer (2), completion (2), versions (2), rejects (2), square (2), balanced (2), invalidate (2), strictly (2), calibrate (2), optimizes (2), quo (2), logistic (2), demonstrating (2), hinder (2), unwarranted (2), evidenced (2), disease (2), opposite (2), suited (2), relationships (2), necessitate (2), respectively (2), characterizations (2), architectures (2), image (2), contribution (2), unify (2), estimable (2), scientific (2), jslz (2), conducting (2), comment (2), statistica (2), sinica (2), logged (2), services (2), individuals (2), compels (2), auditing (2), impossible (2), actual (2), additional (2), monotone (2), audit (2), french (2), identifiability (2), gmm (2), reformulation (2), stakes (2), criminal (2), justice (2), diverse (2), probabilities (2), ability (2), otherwise (2), boundedness (2), receiving (2), regularizing (2), sharing (2), scheme (2), identifiable (2), agree (2), personalizing (2), lack (2), unwieldy (2), poorly (2), unrelated (2), separately (2), shrunken (2), decoupling (2), chain (2), credit (2), policymaking (2), lending (2), incorporate (2), computing (2), blog (2), root (2), markovian (2), precisely (2), hope (2), breaking (2), hölder (2), extremes (2), regions (2), adjusts (2), establishing (2), recovering (2), light (2), dml (2), stands (2), needs (2), bad (2), solves (2), highlighted (2), developments (2), update (2), trivial (2), exhibits (2), batch (2), transitions (2), rendering (2), express (2), nonconvex (2), approximations (2), act (2), stages (2), build (2), achievable (2), itself (2), attain (2), biometrika (2), wherein (2), length (2), dilemma (2), resolve (2), loggers (2), mostly (2), interaction (2), exploitation (2), guaranteed (2), moderate (2), revision (2), subsidies (2), scarce (2), ill (2), turn (2), rejoinder (2), permutation (2), symmetric (2), side (2), msod (2), uniformity (2), appropriately (2), detect (2), morgan (2), therein (2), theorem (2), neuroscience (2), fusion (2), ast (2), exponentiation (2), pointwise (2), levels (2), imply (2), mathematics (2), swaminathan (2), variates (2), exogenous (2), resource (2), utilize (2), perturbation (2), extensively (2), close (2), relationship (2), consideration (2), naive (2), enterprise (2), accepted (2), uniqueness (2), themselves (2), employed (2), answer (2), existence (2), prl (2), erm (2), maximal (2), right (2), investigation (2), validates (2), chambaz (2), dimakopoulou (2), chance (2), want (2), addressed (2)
Text of the page (random words):
allenges our method enables policy learning while simultaneously providing high confidence guarantees using the entire dataset avoiding the need for data splitting we present finite sample and asymptotic versions of our algorithm that ensure the recommended policy satisfies high probability guarantees for avoiding guardrail regressions and or achieving goal outcome improvements we test both variants of our approach approach empirically on a real world application of personalizing sms delivery our results on real world data suggest that our approach offers dramatic improvements in settings with large policy classes and low signal to noise across both finite sample and asymptotic safety guarantees offering up to 300 improvements in detection rates and 150 improvements in policy gains at significantly smaller sample sizes c optimization and personalization causal inference sequential decision making applied ml 0 cspi mt calibrated safe policy improvement with multiple testing for threshold policies with b cho a pop k gan s corbett davies i nir and a evnine acm sigkdd conference on knowledge discovery and data mining kdd 2025 arxiv august 2024 abstract when modifying existing policies in high risk settings it is often necessary to ensure with high certainty that the newly proposed policy improves upon a baseline such as the status quo in this work we consider the problem of safe policy improvement where one only adopts a new policy if it is deemed to be better than the specified baseline with at least pre specified probability we focus on threshold policies a ubiquitous class of policies with applications in economics healthcare and digital advertising existing methods rely on potentially underpowered safety checks and limit the opportunities for finding safe improvements so too often they must revert to the baseline to maintain safety we overcome these issues by leveraging the most powerful safety test in the asymptotic regime and allowing for multiple candidates to be tested for improvement over the baseline we show that in adversarial settings our approach controls the rate of adopting a policy worse than the baseline to the pre specified error level even in moderate sample sizes we present cspi and cspi mt two novel heuristics for selecting cutoff s to maximize the policy improvement from baseline we demonstrate through both synthetic and external datasets that our approaches improve both the detection rates of safe policies and the realized improvement particularly under stringent safety requirements and low signal to noise conditions c optimization and personalization applied ml 0 does weighting improve matrix factorization for recommender systems with a ayoub s robertson d liang and h steck acm web conference www 2025 arxiv october 2025 abstract matrix factorization is a widely used approach for top n recommendations and collaborative filtering when it is implemented on implicit feedback data such as clicks a common heuristic is to upweight the observed interactions this strategy has been shown to improve the performance of certain algorithms in this paper we conduct a systematic study of various weighting schemes and matrix factorization algorithms somewhat surprisingly we find that the best performing methods as measured by the standard unweighted ranking accuracy on publicly available datasets are trained using unweighted data this observation challenges the conventional wisdom in the literature nevertheless we identify cases where weighting can be beneficial particularly for models with lower capacity and certain regularization schemes we also derive efficient algorithms for minimizing a number of weighted objectives which were previously unexplored due to the lack of efficient optimization techniques our work provides a comprehensive analysis of the interplay between weighting regularization and model capacity in matrix factorization for recommender systems c optimization and personalization applied ml 0 collaborative retrieval for large language model based conversational recommender systems with y zhu c wan h steck d liang y feng and j li acm web conference www 2025 arxiv february 2025 abstract conversational recommender systems crs aim to provide personalized recommendations via interactive dialogues with users while large language models llms enhance crs with their superior understanding of context based user preferences they typically struggle to leverage behavioral data which has proven to be the key for classical collaborative filtering approaches for this reason we propose crag collaborative retrieval augmented generation for llm based crs to the best of our knowledge crag is the first approach that combines state of the art llms with collaborative filtering for conversational recommendations our experiments on two publicly available conversational datasets in the movie domain i e a refined reddit dataset as well as the redial dataset demonstrate the superior item coverage and recommendation performance of crag compared to several crs baselines moreover we observe that the improvements are mainly due to better recommendation accuracy on recently released movies j optimization and personalization causal inference 0 adjusting regression models for conditional uncertainty calibration with r gao m yin and j mcinerney machine learning 2024 special issue on conformal prediction and distribution free uncertainty quantification arxiv september 2024 abstract conformal prediction methods have finite sample distribution free marginal coverage guarantees however they generally do not offer conditional coverage guarantees which can be important for high stakes decisions in this paper we propose a novel algorithm to train a regression function to improve the conditional coverage after applying the split conformal prediction procedure we establish an upper bound for the miscoverage gap between the conditional coverage and the nominal coverage rate and propose an end to end algorithm to control this upper bound we demonstrate the efficacy of our method empirically on synthetic and real world datasets j c causal inference 0 inference on strongly identified functionals of weakly identified functions with a bennett x mao w newey v syrgkanis and m uehara journal of the royal statistical society series b jrss b 2025 conference on learning theory colt 2023 extended abstract arxiv august 2022 abstract in a variety of applications including nonparametric instrumental variable npiv analysis proximal causal inference under unmeasured confounding and missing not at random data with shadow variables we are interested in inference on a continuous linear functional e g average causal effects of nuisance function e g npiv regression defined by conditional moment restrictions these nuisance functions are generally weakly identified in that the conditional moment restrictions can be severely ill posed as well as admit multiple solutions this is sometimes resolved by imposing strong conditions that imply the function can be estimated at rates that make inference on the functional possible in this paper we study a novel condition for the functional to be strongly identified even when the nuisance function is not that is the functional is amenable to asymptotically normal estimation at n rates the condition implies the existence of debiasing nuisance functions and we propose penalized minimax estimators for both the primary and debiasing nuisance functions the proposed nuisance estimators can accommodate flexible function classes and importantly they can converge to fixed limits determined by the penalization regardless of the identifiability of the nuisances we use the penalized nuisance estimators to form a debiased estimator for the functional of interest and prove its asymptotic normality under generic high level conditions which provide for asymptotically valid confidence intervals we also illustrate our method in a novel partially linear proximal causal inference problem and a partially linear instrumental variable regression problem c optimization and personalization sequential decision making 0 switching the loss reduces the cost in offline reinforcement learning with a ayoub k wang v liu s robertson j mcinerney d liang and c szepesvari international conference on machine learning icml 2024 arxiv march 2024 abstract we propose training fitted q iteration with log loss fqi log for offline reinforcement learning rl we show that the number of samples needed to learn a near optimal policy with fqi log scales with the accumulated cost of the optimal policy which is zero in problems where acting optimally achieves the goal and incurs no cost in doing so we provide a general framework for proving small cost bounds i e bounds that scale with the optimal achievable cost in offline rl moreover we empirically verify that fqi log uses fewer samples than fqi trained with squared loss on problems where the optimal policy reliably achieves the goal c optimization and personalization sequential decision making 0 more benefits of being distributional second order bounds for reinforcement learning with k wang o oertell a agarwal and w sun international conference on machine learning icml 2024 arxiv february 2024 abstract in this paper we prove that distributional reinforcement learning distrl which learns the return distribution can obtain second order bounds in both online and offline rl in general settings with function approximation second order bounds are instance dependent bounds that scale with the variance of return which we prove are tighter than the previously known small loss bounds of distributional rl to the best of our knowledge our results are the first second order bounds for low rank mdps and for offline rl when specializing to contextual bandits one step rl problem we show that a distributional learning based optimism algorithm achieves a second order worst case regret bound and a second order gap dependent bound simultaneously we also empirically demonstrate the benefit of distrl in contextual bandits on real world datasets we highlight that our analysis with distrl is relatively simple follows the general framework of optimism in the face of uncertainty and does not require weighted regression our results suggest that distrl is a promising framework for obtaining second order bounds in general rl settings thus further reinforcing the benefits of distrl c causal inference sequential decision making 0 peeking with peak sequential nonparametric composite hypothesis tests for means of multiple data streams with b cho and k gan international conference on machine learning icml 2024 arxiv february 2024 abstract we propose a novel nonparametric sequential test for composite hypotheses for means of multiple data streams our proposed method peeking with expectation based averaged capital peak builds upon the testing as betting framework and provides a non asymptotic α level test across any stopping time peak is computationally tractable and efficiently rejects hypotheses that are incorrect across all potential distributions that satisfy our nonparametric assumption enabling joint composite hypothesis testing on multiple streams of data we numerically validate our theoretical findings under the best arm identification and threshold identification in the bandit setting illustrating the computational efficiency of our method against state of the art testing methods c causal inference sequential decision making 0 inferring the long term causal effects of long term treatments from short term experiments with a tran and a bibaut international conference on machine learning icml 2024 arxiv november 2023 oral at icml abstract we study inference on the long term causal effect of a continual exposure to a novel intervention which we term a long term treatment based on an experiment involving only short term observations key examples include the long term health effects of regularly taken medicine or of environmental hazards and the long term effects on users of changes to an online platform this stands in contrast to short term treatments or shocks whose long term effect can reasonably be mediated by short term observations enabling the use of surrogate methods long term treatments by definition have direct effects on long term outcomes via continual exposure so surrogacy cannot reasonably hold our approach instead learns long term temporal dynamics directly from short term experimental data assuming that the initial dynamics observed persist but avoiding the need for both surrogacy assumptions and auxiliary data with long term observations we connect the problem with offline reinforcement learning leveraging doubly robust estimators to estimate long term causal effects for long term treatments and construct confidence intervals finally we demonstrate the method in simulated experiments j optimization and personalization causal inference 0 doubly valid doubly sharp sensitivity analysis for causal inference with unmeasured confounding with j dorn and k guo journal of the american statistical association jasa 2024 arxiv december 2021 github abstract we study the problem of constructing bounds on the average treatment effect in the presence of unobserved confounding under the marginal sensitivity model of tan 2006 combining an existing characterization involving adversarial propensity scores with a new distributionally robust characterization of the problem we propose novel estimators of these bounds that we call doubly valid doubly sharp dvds estimators double sharpness corresponds to the fact that dvds estimators consistently estimate the tightest possible i e sharp bounds implied by the sensitivity model even when one of two nuisance parameters is misspecified and achieve semiparametric efficiency when all nuisance parameters are suitably consistent double validity is an entirely new property for partial identification dvds estimators still provide valid though not sharp bounds even when most nuisance parameters are misspecified in fact even in cases when dvds point estimates fail to be asymptotically normal standard wald confidence intervals may remain valid in the case of binary outcomes the dvds estimators are particularly convenient and possesses a closed form expression in terms of the outcome regression and propensity score we demonstrate the dvds estimators in a simulation study as well as a case study of right heart catheterization c causal inference 0 learning the covariance of treatment effects across many weak experiments with a bibaut w chou and s ejdemyr acm sigkdd conference on knowledge discovery and data mining kdd 2024 arxiv february 2024 github netflix technology blog abstract when primary objectives are insensitive or delayed experimenters may instead focus on proxy metrics derived from secondary outcomes for example technology companies often infer long term impacts of product interventions from their effects on weighted...
|
