The paper studies how noisy labels impact decision-making in machine learning.
problem The impact of noisy labels on decision-making in machine learning.
method Introducing a notion of regret, studying standard approaches, and estimating individual-level mistakes.
result Standard approaches can lead to unforeseen mistakes for individuals, revealing the need for anticipation.
Ahpatron improves online kernel learning with tighter mistake bounds.
problem Improving mistake bounds in online kernel learning with budget constraints.
method Introducing Ahpatron, a new model that uses an aggressive updating rule and a budget maintenance mechanism to approximate AVP.
result Ahpatron achieves tighter mistake bounds compared to previous models.
A deterministic apple tasting learner is developed, confirming a conjecture and providing tight bounds for mistake bounds.
problem Determining the learnability of hypothesis classes in binary online classification with apple tasting feedback.
method Developed a deterministic apple tasting learner and proved tight bounds for mistake bounds.
result Deterministic apple tasting is feasible and provides tight bounds for mistake bounds.
This note corrects one serious mistake and several smaller mistakes from arXiv:math/0502404. The main results of that paper are unchanged.
We retract the scalar curvature rigidity theorem as there is a mistake in the proof. We thank S. Montiel for pointing out the mistake.
For a number of reasons, computational intelligence and machine learning methods have been largely dismissed by the professional community. The reasons for this are numerous and varied, but inevitably amongst the reasons given is that the systems designed often do not perform as expected by their designers. The reasons…
Improved mistake bounds for transductive online learning.
problem Quantifying the power of unlabeled data in online learning.
method Proving lower and upper bounds on transductive mistake bounds.
result Exponential improvement in mistake bounds for transductive learning.
Simplifies online learning with consistent oracle to fewer mistakes.
problem Online learning with computationally intractable Littlestone dimension computation.
method Novel algorithm making at most O ( 256 d ) O(256^d) O ( 25 6 d ) mistakes, simpler proof. result No algorithm can make less than 3 d 3^d 3 d mistakes. Self-directed learners can minimize mistakes in online classification.
problem Minimizing mistakes in online classification with adaptive prediction order.
method Designing efficient self-directed learners for linear classification.
result Strong separation between worst-order and random-order learning for linear classification.
Study the tradeoffs of bandit feedback in multiclass classification.
problem The price of using bandit feedback in multiclass classification.
method Mistake bound model, analysis of variants, and comparison of learners and adversaries.
result The optimal mistake bound under bandit feedback is at most O ( k ) O(k) O ( k ) times higher than in full information, with a tight bound of O ( k ) O(k) O ( k ) . Improved mistake bound for group linear separable cases in online multiclass linear classification.
problem Improving mistake bounds for online multiclass linear classification under group linear separable conditions.
method Refined group weak linear separability condition and rational kernel approach.
result Achieved a mistake bound of K ⋅ 2 i l d e O ( 1 / γ log L ) ) K\cdot 2^{ ilde{O}(\sqrt{1/γ}\log L)}) K ⋅ 2 i l d e O ( 1/ γ l o g L ) ) under group weak linear separable condition. Study on tradeoffs between mistakes and ERM oracle calls in online and transductive learning.
problem Analyzing online and transductive learning with limited ERM and weak consistency oracle access.
method Proves lower bounds and upper bounds on mistakes and oracle calls, considering realizable and agnostic cases.
result Achieves optimal mistake bounds with weak consistency queries for certain concept classes.
SkewSize detects model biases by analyzing mistakes across subgroups.
problem Benchmarking model performance in the presence of spurious correlations.
method Introducing SkewSize, a metric that captures bias from model mistakes.
result SkewSize highlights biases not captured by other metrics.
Thompson Sampling is at most twice as bad as any other policy in Bayesian bandit models.
problem Optimizing selection of the best arm in Bayesian bandit models with independent latent processes.
method Thompson Sampling approach applied to models with independent latent arm processes.
result Thompson Sampling makes at most twice the expected number of mistakes compared to any other policy.
Corrected a mistake in a paper about minimal surfaces.
problem A mistake in a paper about complete minimal surfaces with finite total curvature.
method No new method introduced, just correcting an error.
result Corrected a mistake in a previously published paper.
Using Jeff Holman's comments in Quantitative Finance to illustrate 4 critical errors students should learn to avoid: 1) Mistaking tails (4th moment) for volatility (2nd moment), 2) Missing Jensen's Inequality, 3) Analyzing the hedging wihout the underlying, 4) The necessity of a numeraire in finance.
Open problem seeks an online learning algorithm for binary classification.
problem Existence of an online learning algorithm for binary classification with sublinear mistakes.
method Assumption of sequence allowing learning algorithm's existence.
result Specific condition determines sequence's learnability.
Modified Perceptron handles strategic agents with limited position changes.
problem Learning linear classifiers in the presence of strategic agents that can manipulate their positions.
method Developed a modified Perceptron algorithm with bounded mistakes under various manipulation costs.
result The modified Perceptron achieves bounded mistakes even when manipulation costs are unknown.
Paper analyzes mistake and generalization of MNIC classifiers.
problem Understanding the performance of interpolating classifiers.
method Elementary analyses of MNIC's regret and generalization.
result MNIC generalizes with a rate proportional to the norm of the interpolating solution and inversely proportional to the number of data points.
We investigate the problem of active learning on a given tree whose nodes are assigned binary labels in an adversarial way. Inspired by recent results by Guillory and Bilmes, we characterize (up to constant factors) the optimal placement of queries so to minimize the mistakes made on the non-queried nodes. Our query se…
Study apple tasting feedback in online binary classification, providing new insights into minimax expected mistakes.
problem Online binary classification with partial feedback (apple tasting).
method Combinatorial analysis, Littlestone dimension, Effective width.
result Established a trichotomy of minimax expected mistakes in the realizable setting.
LEAK learns from mistakes to improve point cloud segmentation.
problem Improving point cloud semantic segmentation performance.
method Coarse-to-fine clustering, class-conditional prototypical feature alignment, fairness weighting.
result State-of-the-art performances on different architectures, datasets, and tasks.
Study on computable online learning with new conditions and complexities.
problem Characterizing optimal online learning under varying optimality requirements.
method Introduced anytime optimal (a-optimal) online learning and explored computational separations.
result Found a computational separation between a-optimal and optimal online learning.
Study online learning of neural networks with margin condition.
problem Online learning of neural networks with sign activation function.
method Characterized margin condition for online learnability, proved mistake bounds, constructed counterexamples.
result Proved optimal mistake bounds and lower bounds for neural networks.
Online learning makes sequence of decisions with partial data arrival where next movement of data is unknown. In this paper, we have presented a new technique as multiple times weight updating that update the weight iteratively forsame instance. The proposed technique analyzed with popular state-of-art algorithms from …
A technique to quickly fix mistakes in neural networks.
problem Fixing model errors in neural networks quickly and without affecting other samples.
method Editable Training, a model-agnostic training technique.
result Effectiveness demonstrated on large-scale image classification and machine translation tasks.
Study on learning to predict dynamical systems without assuming their structure.
problem Learning to predict the next state of a dynamical system with unknown evolution function.
method Defined new combinatorial measures to quantify mistake and regret bounds in realizable and agnostic settings.
result In the realizable setting, the number of mistakes can grow arbitrarily with time.
This note corrects the mistakes in the splicing formulas of the paper "Floer homology and splicing knot complements". The mistakes are the result of the incorrect assumption that for a knot K K K inside a homology sphere Y Y Y , the involution on the knot Floer homology of K K K which corresponds to moving the basepoints by o…
John Morgan and G,Tian pointed out a mistake in the concluding argument for our paper entitled " C 1 C_1 C 1 in [2] is zero", which was recently published in arXiv:1512.02098. We hereby acknowledge this mistake and correct the computation, leading to the conclusion that C 1 C_1 C 1 is non-zero and that their reference [2] does inde…
Traders underestimated risk-free rates, leading to poor investments.
problem Incorrect setting of risk-free rates by traders.
method Analysis of investment decisions and financial models.
result Underestimating risk-free rates led to flawed investment decisions.
This paper has been withdrawn by the authors, due a crucial mistake in Lemma 2
New algorithm tackles multiclass transductive online learning with unbounded labels.
problem Characterizing optimal mistake bound for unbounded label spaces.
method Introducing new combinatorial dimensions (Level-constrained Littlestone and Branching dimensions) to characterize online learnability.
result Established trichotomy of possible minimax rates for unbounded label spaces: Θ ( T ) Θ(T) Θ ( T ) , Θ ( log T ) Θ(\log T) Θ ( log T ) , or Θ ( 1 ) Θ(1) Θ ( 1 ) . We study the problem of efficient online multiclass linear classification with bandit feedback, where all examples belong to one of K K K classes and lie in the d d d -dimensional Euclidean space. Previous works have left open the challenge of designing efficient algorithms with finite mistake bounds when the data is linear…
We study the multiclass online learning problem where a forecaster makes a sequence of predictions using the advice of n n n experts. Our main contribution is to analyze the regime where the best expert makes at most b b b mistakes and to show that when b = o ( log 4 n ) b = o(\log_4{n}) b = o ( log 4 n ) , the expected number of mistakes made by the optima…
Gaptron algorithm reduces mistakes in online multiclass classification.
problem Online multiclass classification with limited information.
method Randomized first-order algorithm exploiting the gap between zero-one loss and surrogate losses.
result First linear time algorithm with O ( K T ) O(K\sqrt{T}) O ( K T ) expected regret. This paper contains a correction of a mistake made in arXiv:1405.1324
This article has been withdrawn due to a mistake which is explained in version 2.
Survey of algorithms to correct past mistakes in prediction.
problem Improving prediction accuracy by correcting past errors.
method Defensive Forecasting as a sequential game theory approach to minimize prediction metrics.
result Simple, near-optimal algorithms for various prediction tasks.
This paper has been withdrawn by the authors, due to a mistake pointed out by Lenny Ng and Josh Sabloff.
This paper has been withdrawn by the author due to a mistake in the section 4.
Paper proposes online algorithms for multiclass classification with partial labels.
problem Classifying data with partial labels.
method Avg Perceptron, Max Perceptron, Avg Pegasos, Max Pegasos algorithms.
result Mistake bounds for Avg Perceptron and regret bound for Avg Pegasos.
This paper has some inconsistent results, i.e., we made some failed claims because we did some mistakes for using the test criterion for a series. Precisely, our claims on the convergence rate of O ( 1 / t ) \mathcal{O}(1/t) O ( 1/ t ) of SGD presented in Theorem 1, Corollary 1, Theorem 2 and Corollary 2 are wrongly derived because they ar…
This paper has been withdrawn since it is identical to the paper math.QA/0601267. It was posted by mistake.
The purpose of this erratum is to correct a mistake in the proof of Theorem 4.1 of our paper \cite{CF}.
We point out a mistake in the main statement of \cite{liu} and suggest and proof a correct statement.
We correct a mistake on the citation of JSJ theory in \cite{Ni}. Some arguments in \cite{Ni} are also slightly modified accordingly.
Machine learning models are vulnerable to adversarial examples: small changes to images can cause computer vision models to make mistakes such as identifying a school bus as an ostrich. However, it is still an open question whether humans are prone to similar mistakes. Here, we address this question by leveraging recen…
Improved Markov models learn from their mistakes and adapt to problem complexity.
problem Limitations of standard masked discrete diffusion models in reasoning tasks.
method Learning a Markov transition kernel trained on its own outputs, allowing remasking and adaptation.
result Significant improvement in solving reasoning problems, especially Sudoku-Extreme and Countdown-4.