Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

Machine Learning

886 missions · 536 completed

The science of systems that learn from data and experience. Its scope runs from the statistical and mathematical foundations of learning, including generalization, expressivity, and computational limits, through the design of learning algorithms, deep learning, reinforcement learning, and probabilistic methods, to the empirical study of large models and the trustworthiness, interpretability, and societal impact of learned systems.

Missions

Open350Completed536All886
🏆Completed
Algorithmic Game TheoryOperations Research·Captain: mikedeng1

Fast Convergence of Regularized Learning in Games III: In a (λ,µ)-Smooth Cost Game, First-Order Regret Bounds Make the Average Cost at Most λ(1+µ)/(µ(1−µ))·OPT′ + A·n log d/TResearch Paper

When independent players repeatedly choose actions, a low-regret guarantee controls each player's performance against a fixed alternative. It does not directly control the social cost, the sum of all players' costs. For a class of games satisfying smoothness, however, the individual comparisons can imply a bound on the cost of the joint play. This…

8 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsOperations ResearchStatistics·Captain: mikedeng1

Explore First, Exploit Next: The True Shape of Regret in Bandit Problems III: A Strategy Smarter Than Uniform Draws Every Arm at Least (T/K)(1 − √(2T K_inf)) TimesResearch Paper

A stochastic multi-armed bandit learner repeatedly selects one of finitely many arms and sees only the reward from the selected arm. The learner must gather information about uncertain rewards while trying to select high-mean arms. Expected regret summarizes the lost reward relative to always selecting an optimal arm, but regret alone does not explain how…

11 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsReinforcement Learning·Captain: mikedeng1

The Statistical Complexity of Interactive Decision Making V: The Multi-Armed Bandit Has Localized DEC at Least 2⁻⁶·A/γ, via Hard Families of ModelsResearch Paper

The Decision-Estimation Coefficient (DEC) of Foster, Kakade, Qian and Rakhlin (arXiv:2112.13487) is a single complexity measure that bounds the minimax regret of interactive decision making (bandits, reinforcement learning, and their structured variants) from above and from below. Its lower bound (Theorem 3.2 of the paper) states that every algorithm…

13 thms5 active usersReviewed
🏆Completed
Information TheoryStatistics·Captain: mikedeng1

The Statistical Complexity of Interactive Decision Making I: Subadditivity of Squared Hellinger Distance for Sequential Laws, D²_H(P,Q) ≤ 10² log(n)·E_P[Σᵢ D²_H(P⁽ⁱ⁾, Q⁽ⁱ⁾)] for n ≥ 2Research Paper

Lower bounds for interactive decision making (bandits, reinforcement learning, and the general decision making with structured observations framework of Foster, Kakade, Qian and Rakhlin, arXiv:2112.13487) compare what a learner observes under two different environments. A learner's whole interaction history is generated one step at a time: each observation…

23 thms5 active users
🏆Completed
Operations ResearchOptimizationProbability·Captain: mikedeng1

Stochastic First- and Zeroth-Order Methods for Nonconvex Stochastic Programming 2: The Randomized Stochastic Gradient-Free Method with Gaussian Smoothing Bounds (1/L)E‖∇f(x_R)‖² for Nonconvex fResearch Paper

Many optimization problems in simulation-based design, reinforcement learning and machine learning can be evaluated but not differentiated. Their objective is an expectation f(x)=E[F(x,ξ)]f(x) = \mathbb E[F(x,\xi)]f(x)=E[F(x,ξ)] that can only be sampled through noisy function values F(x,ξ)F(x,\xi)F(x,ξ). Gradients may be unavailable because the model is a black box, a simulator, or a legacy code.…

18 thms5 active usersReviewed
🏆Completed
ProbabilityStatistics·Captain: mikedeng1

Distribution-Free, Risk-Controlling Prediction Sets II: The Waudby-Smith–Ramdas Betting Bound Is a (1 − δ) Upper Confidence Bound for the Mean Loss (Proposition 5)Research Paper

A risk-controlling prediction set is a set-valued predictor T\mathcal TT that, with probability at least 1−δ1-\delta1−δ over a calibration sample, has expected loss at most a target level α\alphaα. Bates, Angelopoulos, Lei, Malik and Jordan (arXiv:2101.02703, J. ACM 2021) build such sets for any black-box predictor and any bounded loss by a procedure called upper…

7 thms5 active usersReviewed
🏆Completed
Numerical AnalysisOptimal TransportOptimization·Captain: mikedeng1

Computational Optimal Transport VIII: Sinkhorn's Scalings Contract by λ(K)² per Iteration in Hilbert's Projective MetricTextbook

Sinkhorn's algorithm is the standard method for computing entropically regularized optimal transport between two histograms. It is used throughout machine learning, statistics, imaging and graphics because each iteration is two matrix–vector products followed by entrywise divisions, which parallelize on GPUs and differentiate easily. The same iteration,…

19 thms5 active users
🏆Completed
Operations ResearchOptimal TransportOptimization·Captain: mikedeng1

Regularization via Mass Transportation II: Wasserstein-Robust Linear Regression Equals Empirical Loss plus a Dual-Norm PenaltyResearch Paper

Ordinary least squares and its robust relatives (least absolute deviations, Huber, ϵ\epsilonϵ-insensitive and quantile regression) fit a linear predictor by minimizing an average loss over the training sample. That average estimates the out-of-sample error poorly when the sample is small, and the usual fix is to add a regularization penalty such as a norm of…

15 thms5 active usersReviewed
🏆Completed
Combinatorics·Captain: mikedeng1

Online Learning and Online Convex Optimization 4: Finite Littlestone Dimension Gives Agnostic Online Regret 2√(Ldim·ln(eT/Ldim)·T)Research Paper

Online classification is the simplest model of learning from a stream: on every round a learner sees an instance, guesses its binary label, and is then told the truth. Cover's impossibility argument shows that against an adversary no deterministic learner can compete with even two fixed hypotheses, so two relaxations are studied. In the realizable case the…

8 thms5 active usersReviewed
🏆Completed
ProbabilityTheoretical Computer Science·Captain: mikedeng1

What Can We Learn Privately? II: Parity Functions Are Privately PAC Learnable from O(d log(1/β)/(εα)) ExamplesResearch Paper

Differential privacy (Dwork, McSherry, Nissim and Smith, 2006) requires a randomized algorithm's output distribution to change little when one individual's record is changed. PAC learning asks an algorithm to output, from i.i.d. labeled examples, a hypothesis whose error is small with high probability. Kasiviswanathan, Lee, Nissim, Raskhodnikova and Smith (a…

14 thms5 active usersReviewed
🏆Completed
ProbabilityTheoretical Computer Science·Captain: mikedeng1

What Can We Learn Privately? I: The Exponential Mechanism Privately Agnostically Learns Any Finite Concept Class from O((ln|C| + ln(1/β))·max{1/(εα), 1/α²}) ExamplesResearch Paper

Many learning tasks run on data about people: medical records, purchase histories, survey answers. A learner whose output reveals too much about any one training example is unusable on such data, however accurate it is. Differential privacy (Dwork, McSherry, Nissim and Smith, 2006) makes "reveals too much" precise: changing one individual's record may…

10 thms5 active usersReviewed
🏆Completed
OptimizationStatistics·Captain: mikedeng1

Additive Logistic Regression: A Statistical View of Boosting 2: Population Discrete AdaBoost Is a Newton-Like Step with Exact Line Search c = ½ log((1−err)/err)Research Paper

AdaBoost (Freund and Schapire, 1997) combines many weak classifiers fm(x)∈{−1,1}f_m(x) \in \{-1, 1\}fm​(x)∈{−1,1} into a weighted vote sign⁡(∑mcmfm(x))\operatorname{sign}\big(\sum_m c_m f_m(x)\big)sign(∑m​cm​fm​(x)). At each round it fits a classifier to a reweighted training sample, sets cm=log⁡((1−errm)/errm)c_m = \log((1 - \mathrm{err}_m)/\mathrm{err}_m)cm​=log((1−errm​)/errm​) from the weighted error errm\mathrm{err}_merrm​, and multiplies the weights of the…

10 thms5 active usersReviewed
🏆Completed
OptimizationProbabilityStatistics·Captain: mikedeng1

Learning Models with Uniform Performance via Distributionally Robust Optimization 2: A Minimax Lower Bound of Order n^(−1/2) ∨ n^(−1/k*) for Estimating the Cressie–Read Robust Risk (Theorem 3)Research Paper

Distributionally robust optimization replaces the expected loss EP0[ℓ(θ;X)]\mathbb E_{P_0}[\ell(\theta; X)]EP0​​[ℓ(θ;X)] of a statistical model by its worst case over all distributions QQQ close to the data distribution P0P_0P0​. When closeness is measured by an fff-divergence ball {Q:Df(Q∥P0)≤ρ}\{Q : D_f(Q\|P_0) \le \rho\}{Q:Df​(Q∥P0​)≤ρ} of fixed radius ρ>0\rho > 0ρ>0, the robust risk upper-bounds the risk of every…

13 thms5 active usersReviewed
🏆Completed
Operations ResearchTheoretical Computer Science·Captain: mikedeng1

The Weighted Majority Algorithm 8: With η Anomalies Every Deterministic Learner Can Be Forced to opt(F, 0) + 2η MistakesResearch Paper

In the on-line prediction model studied by Littlestone and Warmuth (Inform. and Comput. 108, 1994), a learner receives instances one at a time, predicts a binary label for each, and is then told the correct label. No probabilistic assumption is made: performance is measured by the number of mistakes on a worst-case sequence. When every label is given by…

6 thms5 active usersReviewed
🏆Completed
Dynamic ProgrammingOperations ResearchProbability+1·Captain: mikedeng1

Stable Function Approximation in Dynamic Programming II: In a Nondiscounted MDP with All Policies Proper, Value Iteration Through a Self-Weighted Averager ConvergesResearch Paper

Value iteration solves a Markov decision process by repeatedly applying the Bellman backup to a table of state values. When the state space is too large for a table, the values are stored in a function approximator instead: each backup is followed by a fit, and the fitted function is backed up again. This approximate value iteration is the basis of fitted…

8 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsOperations ResearchProbability·Captain: mikedeng1

MNL-Bandit: A Dynamic Learning Approach to Assortment Selection I: The Epoch-Based UCB Policy Has Regret at Most C₁√(NT log NT) + C₂N log² NT Under the No-Purchase AssumptionResearch Paper

A retailer that decides which products to display, an online platform that decides which items to show in a recommendation slot, and an airline that decides which fare classes to open all face the same problem: the set of options offered changes what customers buy, and the substitution pattern is not known in advance. The multinomial logit (MNL) model is…

15 thms5 active usersReviewed
🏆Completed
Convex OptimizationOptimizationProbability·Captain: mikedeng1

Convex Optimization: Algorithms and Complexity XV: SVRG with η = 1/(10β) and k = 20κ Contracts the Expected Optimality Gap by 0.9 per EpochTextbook

Many optimization problems in machine learning minimize an average of losses, one loss for each observation. A full gradient step examines every observation, while a stochastic gradient step examines one. The latter is cheaper per step, but its sampled gradient can remain noisy even near the optimum. Section 6.3 of Bubeck's monograph studies stochastic…

14 thms5 active usersReviewed
🏆Completed
Convex OptimizationOptimization·Captain: mikedeng1

Convex Optimization: Algorithms and Complexity III: Projected Subgradient Descent with η = R/(L√t) Satisfies f(average) − f(x*) ≤ RL/√tTextbook

Many convex optimization problems in machine learning and statistics have objectives that are convex but not differentiable: hinge losses, ℓ1\ell_1ℓ1​ penalties, maxima of finitely many affine functions, and the dual functions of Lagrangian relaxations. Methods that rely on gradients do not apply to them directly, while cutting-plane methods such as the…

15 thms5 active users
🏆Completed
ProbabilityStatistics·Captain: mikedeng1

Rademacher and Gaussian Complexities: Risk Bounds and Structural Results 3: Monotone, Absolutely Homogeneous, Convex-Hull Invariant and Subadditive R_n, and the 2L Bound for L-Lipschitz Maps Fixing 0Research Paper

Generalization bounds in statistical learning theory control the gap between the empirical and the true performance of a predictor chosen from a class FFF. Bartlett and Mendelson (JMLR 3, 2002) showed that this gap is governed by the Rademacher complexity of a loss class built from FFF, and that, unlike the VC dimension, this complexity can be estimated from…

7 thms5 active usersReviewed
🏆Completed
ProbabilityStatistics·Captain: mikedeng1

Rademacher and Gaussian Complexities: Risk Bounds and Structural Results 1: With Probability 1 − δ, Expected Loss ≤ Empirical Dominating Cost + R_n(φ̃∘F) + √(8 ln(2/δ)/n) for Every f in FResearch Paper

A learning algorithm picks a predictor fff from a class FFF after looking at nnn training examples, and the quantity of interest is its expected loss on a fresh example. Since fff depends on the data, its training error is a biased estimate of that loss, and a risk bound quantifies the bias uniformly over the class: with high probability, every f∈Ff \in Ff∈F has expected…

10 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsOperations ResearchProbability·Captain: mikedeng1

Analysis of Thompson Sampling for the Multi-armed Bandit Problem 1: Logarithmic Regret for Two ArmsResearch Paper

Thompson Sampling (TS) is the oldest heuristic for the stochastic multi-armed bandit problem: it was proposed by Thompson in 1933 (Biometrika 25) and is used in practice for online advertising and recommendation, where it often performs as well as or better than upper-confidence-bound methods (Chapelle and Li, NIPS 2011; Scott 2010). Until 2012 its…

12 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsOperations ResearchProbability+1·Captain: mikedeng1

Optimal Best Arm Identification with Fixed Confidence IV: Asymptotic Optimality of the Track-and-Stop StrategyResearch Paper

In best arm identification with fixed confidence, a learner samples KKK unknown distributions (arms) sequentially and must, as early as possible, name the arm with the largest mean, while being wrong with probability at most a prescribed risk δ\deltaδ. The problem models adaptive A/B/n testing, clinical and simulation-based selection among alternatives, and the…

17 thms5 active usersReviewed
🏆Completed
Bandit AlgorithmsOperations ResearchProbability+1·Captain: mikedeng1

Optimal Best Arm Identification with Fixed Confidence I: Non-Asymptotic Lower Bound on the Sample ComplexityResearch Paper

Best arm identification with fixed confidence is the pure-exploration counterpart of the multi-armed bandit problem. A learner faces KKK unknown reward distributions ("arms"), samples them sequentially, and must stop and name the arm with the largest mean, being wrong with probability at most a prescribed δ\deltaδ. The question is how many samples this…

15 thms5 active usersReviewed
🏆Completed
ProbabilityStatistics·Captain: mikedeng1

On the Uniform Convergence of Relative Frequencies of Events to Their Probabilities III: Uniform Convergence and the Entropy per ObservationResearch Paper

Estimating a probability by the relative frequency of the event in an independent sample is justified for one event by the law of large numbers. Statistics and learning theory need more: the frequencies of a whole class of events SSS must approach their probabilities simultaneously, so that a quantity chosen after looking at the data (the empirical risk…

20 thms5 active usersReviewed
🏆Completed
Convex OptimizationProbabilityRandom Matrix Theory+1·Captain: mikedeng1

The Power of Convex Relaxation: Near-Optimal Matrix Completion II: Exact Nuclear-Norm Recovery from Nearly Minimally Many EntriesResearch Paper

Many data sets are large matrices of which only a small fraction of the entries is observed, and of which the underlying object is believed to have low rank: user–item rating tables in collaborative filtering, distance matrices in sensor-network localization, and measurement matrices in structure-from-motion. Matrix completion asks when the missing entries…

22 thms5 active usersReviewed
PreviousPage 4 of 36Next

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me