Study shows flipping a small subset of labels can severely damage machine learning models.
problem Adversarial attacks on distributed machine learning models.
method Formalized label flipping attacks, proposed a greedy algorithm, demonstrated with logistic regression models.
result A budget of only 0.1% of labels at each training step can reduce model accuracy by 6%, and some models can perform worse than random guessing when up to 25% of labels are flipped.
Efficiently poisons offline RLHF models by flipping preference labels.
problem Vulnerability of offline RLHF models to preference label flipping attacks.
method Developed two attack methods: BAL-A and BMP-A, solving a structured binary sparse approximation problem.
result Demonstrated that flipping one preference label induces a parameter-independent shift in the DPO gradient, enabling structured binary sparse approximation.
Deep Partition Aggregation defends against poisoning attacks with provable certificates.
problem Adversarial poisoning attacks corrupt classifier test-time behavior.
method Deep Partition Aggregation (DPA) is an ensemble method using hash partitions and base models trained on these partitions.
result DPA can certify >= 50% of test images against over 500 poison image insertions on MNIST, and nine insertions on CIFAR-10.
Many machine learning systems rely on data collected in the wild from untrusted sources, exposing the learning algorithms to data poisoning. Attackers can inject malicious data in the training dataset to subvert the learning process, compromising the performance of the algorithm producing errors in a targeted or an ind…
New method makes machine learning models robust to label flipping attacks.
problem Machine learning models are vulnerable to label flipping attacks.
method Randomized smoothing over arbitrary functions to build certifiably robust classifiers.
result Linear classifiers are robust to label flipping attacks with deterministic bounds.
A double pants decomposition of a 2-dimensional surface is a collection of two pants decomposition of this surface introduced in arXiv:1005.0073v2. There are two natural operations acting on double pants decompositions: flips and handle twists. It is shown in arXiv:1005.0073v2 that the groupoid generated by flips and h…
Paper defends against malware detection attacks using clustering and deep learning.
problem Label flipping attacks on malware detection systems in IoT environments.
method Developed a Silhouette clustering-based attack mechanism and two CNN-type deep learning algorithms for defense.
result Proposed algorithms LSD and CSD improve malware detection accuracy by up to 19%.
The paper solves pentagon equations using triangulations and edge transformations.
problem Solving pentagon equations with triangulations and edge transformations.
method General data and transformation rule method applied to triangulations.
result Recovery of initial data after transformations.
It has been a long-standing problem to efficiently learn a halfspace using as few labels as possible in the presence of noise. In this work, we propose an efficient Perceptron-based algorithm for actively learning homogeneous halfspaces under the uniform distribution over the unit sphere. Under the bounded noise condit…
Federated learning is protected against adversarial attacks with residual-based reweighting.
problem Adversarial attacks on federated learning's aggregation process.
method Residual-based reweighting combined with iteratively reweighted least squares.
result Our aggregation algorithm outperforms other methods in label-flipping and backdoor attacks.
New insights on robust learning under strong noise models.
problem Challenging label-noise models in robust learning.
method Extending statistical query framework to more general noise models and using evolutionary algorithms.
result First polynomial time algorithm for learning linear threshold functions with arbitrarily small excess error in presence of Tsybakov noise.
The paper explores how supervision level affects both statistical accuracy and computational efficiency in weakly supervised binary classification.
problem The impact of label flip probability on statistical and computational efficiency in weakly supervised binary classification.
method Information-theoretic and computational boundaries were established to characterize the relationship between supervision level and performance.
result The gap between statistical and computational boundaries narrows as the supervision level increases, indicating improved computational efficiency with more supervision.
Study flip graphs for surfaces of infinite type, finding uncountably many connected components.
problem Understanding relationships between triangulations of infinite type surfaces via flips.
method Associate triangulations to flip graphs and study sequences of simultaneous flips.
result Flip graphs for infinite type surfaces have uncountably many connected components.
Instance- and Label-dependent label Noise (ILN) widely exists in real-world datasets but has been rarely studied. In this paper, we focus on Bounded Instance- and Label-dependent label Noise (BILN), a particular case of ILN where the label noise rates -- the probabilities that the true labels of examples flip into the …
Proposes a method to increase diversity without sacrificing meritocracy.
problem Systemic bias in datasets affecting diversity and meritocracy.
method Optimally flipping outcome labels and training classification models simultaneously.
result The price of diversity is low and sometimes negative, enhancing diversity without significantly affecting meritocracy.
Two-layer ReLU networks can overfit without harm, study finds.
problem Understanding when and how two-layer ReLU networks can overfit without harming generalization.
method Established algorithm-dependent risk bounds for two-layer ReLU convolutional neural networks with label-flipping noise.
result Gradient descent-trained ReLU networks can achieve near-zero training loss and Bayes optimal test risk.
Finite subgraphs in flip graphs ensure unique surface embeddings.
problem Ensuring unique embeddings of surfaces based on flip graphs.
method Analyzing finite subgraphs within flip graphs of surfaces.
result Injective homomorphisms are uniquely extendable and induced by embeddings.
New algorithm uses conditionally invariant components to improve domain adaptation performance.
problem Improving domain adaptation performance when source and target data distributions differ.
method Conditionally invariant components (CICs) and importance-weighted conditional invariant penalty (IW-CIP) algorithm.
result New algorithm provides target risk guarantees and addresses label-flipping features.
A new loss function α-loss improves classification robustness and calibration.
problem Improving classification robustness and calibration in machine learning.
method Introduces a tunable loss function α-loss, parameterized by α, and analyzes its theoretical and practical properties. result The α-loss function can improve model robustness to label flips and sensitivity to imbalanced classes. Study of flip graphs and their automorphism groups for infinite-type surfaces.
problem Understanding automorphism groups of flip graphs for infinite-type surfaces.
method Examined the relationship between mapping class groups and flip graphs for infinite-type surfaces.
result Extended mapping class groups are isomorphic to proper subgroups of automorphism groups of flip graphs.
Connected flip graphs for triangulations on hyperbolic surfaces.
problem Connecting triangulations on hyperbolic surfaces via flips.
method Proving connectedness of flip graphs and giving bounds on edge flips.
result Flip graphs of geometric triangulations are connected.
DCGANs generate realistic breast masses for mammography.
problem Lack of labelled data and imbalanced datasets in medical imaging.
method DCGANs trained on mammographic images to generate synthetic masses.
result DCGANs improve lesion detection in mammography by ~0.09 F1 score.
Study of skateboard flips as continuous curves in SO(3) group.
problem Characterize skateboard flip tricks as continuous motions.
method Model flips as curves in SO(3), analyze lifts to S3, derive formulas. result There are only four distinct flip tricks up to continuous deformation.
Neural networks have been criticized for their lack of easy interpretation, which undermines confidence in their use for important applications. Here, we introduce a novel technique, interpreting a trained neural network by investigating its flip points. A flip point is any point that lies on the boundary between two o…
Flip symmetry on knot diagrams affects Khovanov homology.
problem Understanding the flip map on Khovanov homology.
method Analyzing the behavior of the flip map on unlinks and using it to determine the involution.
result The flip map is the identity map over \(\mathbb{F}_2\), confirming a conjecture.
In this paper, we study a classification problem in which sample labels are randomly corrupted. In this scenario, there is an unobservable sample with noise-free labels. However, before being observed, the true labels are independently flipped with a probability ρ∈[0,0.5), and the random label noise can be class-co…
The flip graph and arc complex of a surface are shown to have finite rigidity.
problem Finite rigidity of flip graph and arc complex for surfaces.
method Embedding the flip graph in the arc complex and leveraging finite rigidity of the flip graph.
result Finite rigidity of the flip graph implies finite rigidity of the arc complex.
Novel defense algorithm improves SVMs against data poisoning attacks.
problem Vulnerability of SVMs to targeted training data manipulations like poisoning attacks.
method Developed a weighted SVM using K-LID to de-emphasize suspicious data samples.
result Significant reduction in classification error rates (10% on average) with the proposed defense.
New examples show flip distance and polyhedron triangulation numbers differ, with ratio close to 3/2.
problem Understanding the relationship between flip distance and polyhedron triangulation numbers.
method Provided examples to demonstrate the difference between flip distance and polyhedron triangulation numbers.
result Ratio of flip distance to polyhedron triangulation numbers can be arbitrarily close to 3/2.
We prove that every injective simplicial map F(S)→F(S′) between flip graphs is induced by a subsurface inclusion S→S′, except in finitely many cases. This extends a result of Korkmaz--Papadopoulos which asserts that every automorphism of the flip graph of a surface without boundary is ind…
Paper proposes training deep nets on noisy labels without manual annotation.
problem Training deep neural networks on noisy labels without manual annotation.
method Directly train deep neural network on noisy candidate labels, early stopping to avoid overfitting.
result Training on noisy candidate labels yields higher test performance than on clean data.
We prove that for a given flat surface with conical singularities, any pair of geometric triangulations can be connected by a chain of flips.
We introduce a notion of cross-flips: local moves that transform a balanced (i.e., properly (d+1)-colored) triangulation of a combinatorial d-manifold into another balanced triangulation. These moves form a natural analog of bistellar flips (also known as Pachner moves). Specifically, we establish the following the…
This paper is about the geometry of flip-graphs associated to triangulations of surfaces. More precisely, we consider a topological surface with a privileged boundary curve and study the spaces of its triangulations with n vertices on the boundary curve. The surfaces we consider topologically fill this boundary curve s…
Geodesics count exponentially between triangulations of surfaces with enough topology.
problem Counting geodesics in triangulations of surfaces.
method Analyzing the flip-graph of triangulations and their geodesics.
result The number of geodesics grows exponentially for surfaces with enough topology.
A method to learn transition matrices without anchor points improves classifier performance.
problem Learning transition matrices in label-noise learning without anchor points.
method Transition-revision (T-Revision) method to learn transition matrices. result The proposed method leads to better classifiers without anchor points.
Identifies minimal training subset to flip a prediction.
problem Flipping predictions in machine learning models.
method Extended influence function for relabeling minimal subset.
result Relabeling fewer than 2% of training points can flip a prediction.
New method improves LLM judge accuracy by accounting for dependencies in aggregated binary labels.
problem Classical label aggregation methods fail to account for dependencies among LLM judges, leading to miscalibrated predictions.
method Dependence-aware models based on Ising graphical models and latent factors.
result The proposed method outperforms classical methods on real-world datasets, reducing excess risk.
In order to model volatile real-world network behavior, we analyze phase-flipping dynamical scale-free network in which nodes and links fail and recover. We investigate how stochasticity in a parameter governing the recovery process affects phase-flipping dynamics, and find the probability that no more than q% of nodes…
Let Σ be a compact surface. We prove that the set of surface cubications modulo flips, up to isotopy, is in one-to-one correspondence with Z/2Z⊕H1(Σ,Z/2Z).
We use flip points to explain and audit deep learning models, revealing decision boundaries and improving model performance.
problem Lack of interpretability in deep learning models hinders their use in important applications.
method Flip points are used to analyze decision boundaries of deep learning models with continuous output scores.
result Flip points reveal the least changes in input that would alter a model's classification, enabling better understanding and improvement of model behavior.
Study finds flipped classrooms improve student self-concept, enjoyment, but not exam scores.
problem Evaluating the impact of flipped classrooms on higher education outcomes.
method Double/debiased machine learning (DML) approach to analyze student data.
result No significant positive effects on exam scores, passing rates, or knowledge retention.
Study quasisymmetric maps on hyperbolic plane boundaries.
problem Identify quasisymmetric maps corresponding to specific lambda lengths and flip distances.
method Analyze maps on Farey triangulation, relate to shearing coordinates and flip distance.
result Identify quasisymmetric maps corresponding to pinched lambda lengths and flip distances.
We study flip-graphs of triangulations on topological surfaces where distance is measured by counting the number of necessary flip operations between two triangulations. We focus on surfaces of positive genus g with a single boundary curve and n marked points on this curve; we consider triangulations up to homeomor…
Ridge regression shows different behaviors in binary classification with noisy labels.
problem Binary classification with noisy labels and anisotropic cluster distributions.
method Investigation of ridge regression behavior in overparameterized settings with label noise.
result Ridge regression exhibits qualitatively different behavior based on the scale of cluster mean vectors and covariance matrices.
Local graph clustering improves with noisy labels, enhancing accuracy and performance.
problem Local graph clustering with noisy labels for node information.
method Constructing a weighted graph with noisy labels and using diffusion-based clustering.
result Diffusion in the weighted graph yields more accurate recovery of target clusters.
Using existing technology, we prove a Masur-Minsky style distance formula for flip- graph distance between two triangulations, expressed as a sum of the distances of the projections of these triangulations into arc graphs of the suitable subsurfaces of S.
Adversarial examples are important for understanding the behavior of neural models, and can improve their robustness through adversarial training. Recent work in natural language processing generated adversarial examples by assuming white-box access to the attacked model, and optimizing the input directly against it (E…