Simple regularization methods mimic adversarial training's robustness.
problem Expensive adversarial training for robustness.
method Label smoothing and logit squeezing.
result Achieves strong adversarial robustness without adversarial examples.
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,742 papers · 148 categories
Trend · papers per month
48 results for “logit squeezing”
Recently, Kannan et al. [2018] proposed several logit regularization methods to improve the adversarial robustness of classifiers. We show that the computationally fast methods they propose - Clean Logit Pairing (CLP) and Logit Squeezing (LSQ) - just make the gradient-based optimization problem of crafting adversarial …
Label smoothing improves model robustness against misspecification.
problem Improving model robustness against model misspecification.
method Introducing modified label smoothing (MLSLR) that maintains consistent probability estimation while modifying the loss function.
result MLSLR exhibits higher robustness against model misspecification than conventional label smoothing.
Extends Gromov non-squeezing to locally conformally symplectic structures.
problem Generalizing Gromov non-squeezing to new geometric structures.
method Deformation theory applied to locally conformally symplectic structures.
result Proves a new extension of the Gromov non-squeezing phenomenon.
Formula for squeezing function on annuli disproves conjecture.
problem Proving squeezing function formula for annuli.
method Schottky-Klein prime function and Loewner differential equation.
result Formula for squeezing function on annuli established.
Weight Squeezing transfers knowledge from large models to smaller ones, improving performance and speed.
problem Transfer learning and model compression for faster and more efficient training.
method Reparameterization of weights from a large model to a smaller one.
result Weight Squeezing outperforms other methods on GLUE benchmark with faster training.
Study non-squeezing phenomena in contact geometry using specific capacities.
problem Detect and quantify non-squeezing in contact geometry.
method Defined and computed two contact capacities, using spectral selectors and Givental's non-linear Maslov index.
result Discovered and quantified non-squeezing phenomena in lens spaces and strongly order able closed prequantizations.
A knot is not squeezable if it fails to meet certain invariant criteria.
problem Proving a specific pretzel knot is not squeezable.
method Comparing Rasmussen invariant and Iida-Taniguchi invariant.
result The knot is not squeezable.
Formulas for tau and epsilon concordance invariants of braided satellite knots
problem tau and epsilon invariants of satellite knots
method tau and epsilon invariants of braided satellite knots
result tau and epsilon formulas for braided satellite knots
New method detects non-product domains using squeezing function.
problem Detecting non-product bounded pseudoconvex domains.
method New application of squeezing function and optimal estimates.
result Identifies new family of holomorphic homogeneous regular domains.
Squeezed Knotsmath.GT
Squeezed knots are slices of minimal cobordisms; obstructions come from quantum knot invariants.
problem Characterizing and obstructing squeezed knots.
method Analysis of cobordisms, quantum knot invariants, and stable cohomology operations.
result Effective obstructions to squeezedness come from quantum knot invariants, notably Rasmussen invariant refinements.
The paper extends Gromov's non-squeezing theorem to deformed symplectic forms.
problem Extending Gromov's non-squeezing theorem to deformed symplectic forms.
method Trap idea for holomorphic curves analogous to dynamical systems.
result The classical Gromov argument breaks down for deformed forms.
Improved non-squeezing theorem for calibrated geometries proved.
problem Proving an improved non-squeezing theorem for calibrated geometries.
method Two proofs: direct and reduction to classical case.
result Established an improved non-squeezing theorem for calibrated geometries.
Paper develops proper, lower-bounded losses for weakly supervised classification.
problem Weakly supervised classification with corrupted labels.
method Representation theorem for proper losses, derived condition for lower-boundedness, generalized logit squeezing.
result Proper and lower-bounded losses for weak-label learning.
Infinitesimal holomorphic realizations for the Schrödinger-Weil representation and the discrete series representations of the Jacobi group are constructed. Explicit expressions of the basic differential operators are obtained. The squeezed states for the unitary irreducible representation of the Jacobi group are introd…
We analyze the relationship between the covering of the Jacobi group and the squeezed states. We attach some nonclassical states to the Jacobi group. The matrix elements of the Jacobi group are presented.
Two new proofs of Gromov's non-squeezing theorem using curve reparametrization and gradient bounds.
problem Gromov's non-squeezing theorem in symplectic geometry.
method Reparametrization of pseudo-holomorphic curves and application of mean value inequality or Gromov-Schwarz lemma.
result Uniform bounds on the gradient of pseudo-holomorphic curves leading to compactness of moduli space.
Logit regularization induces logit clustering, affecting classifier performance.
problem Understanding the mechanism of logit regularization in classification.
method Analysis of logit regularization in linear classification, proving logit clustering leads to Fisher's Linear Discriminant alignment.
result Logit regularization can halve critical sample complexity and induce robust generalization.
Contact squeezing prevented in certain prequantized balls via generating functions.
problem Preventing contact squeezing in prequantized balls of different radii.
method Equivariant generating function homology with finite cyclic group action.
result Contact squeezing not possible in specified prequantized balls.
The study explains how market-makers' hedging affects stock volatility during gamma-squeeze events.
problem Endogenous volatility amplification in option markets during gamma-squeeze events.
method Developed a theoretical framework linking hedging behavior and market turbulence, incorporating beta-normalized volatility.
result Low-beta stocks amplify volatility more during gamma-squeeze events.
Starting from the work of Bhupal, we extend to the contact case the Viterbo capacity and Traynor's construction of symplectic homology. As an application we get a new proof of the Non-Squeezing Theorem of Eliashberg, Kim and Polterovich.
Improves adversarial robustness by constraining logits with a bounded function.
problem Improving adversarial robustness in deep learning models.
method Addition of a bounded function before softmax to constrain logits.
result Our method improves adversarial robustness without requiring adversarial training.
A reinforcement learning approach prepares quantum squeezed states in open spin systems.
problem Generating non-classical states in open quantum systems with dissipation and dephasing.
method Reinforcement learning to determine optimal control pulses for spin-squeezing.
result Optimal control sequences enhance collective spin squeezing and entanglement.
The paper quantifies how much of a 4-ball must be removed to squeeze into a cylinder, proving a lower bound on the Minkowski dimension.
problem Quantifying how much of a 4-ball must be removed to fit into a cylinder.
method Gromov's non-squeezing theorem and Minkowski dimension analysis.
result The Minkowski dimension of the removed set is at least 2, with an example showing this is optimal for certain radii.
Feature Squeezing is a recently proposed defense method which reduces the search space available to an adversary by coalescing samples that correspond to many different feature vectors in the original space into a single sample. It has been shown that feature squeezing defenses can be combined in a joint detection fram…
Logit-GFN accelerates GFlowNets training by scaling logits based on temperature.
problem Training temperature-conditional GFlowNets is numerically challenging.
method Logit-GFN uses a learned function of temperature to scale policy logits.
result Logit-GFN greatly accelerates GFlowNets training and improves generalization and mode discovery.
On Logit Weibull Manifoldmath.ST
A new logit model derived from the Weibull manifold.
problem No potential function on the Weibull manifold.
method Extracted a logit model from the two-parameter Weibull model.
result Found a completely integrable Hamiltonian gradient system on the logit model.
The paper defines and analyzes coherent and squeezed states on manifolds and their quantization.
problem Defining and characterizing coherent and squeezed states on various manifolds.
method Definition and analysis of Rawnsley-type coherent and squeezed states, Berezin quantization.
result Properties and quantization of coherent and squeezed states on manifolds.
Proposes SOVR loss to improve adversarial robustness by increasing logit margins.
problem Adversarial training's difficulty in robustness against sophisticated attacks.
method Introduces SOVR loss function that switches from cross-entropy to one-vs-the-rest loss for important samples.
result SOVR loss increases logit margins of important samples, improving robustness against Auto-Attack.
Let $f : U\subset\Rm \to \calQ_Q(\ell_2)$ be of Sobolev class , . If almost minimizes its Dirichlet energy then is Hölder continuous. If and is squeeze and squash stationary then is in VMO.
The paper studies invariant weighted Bergman metrics on domains.
problem Investigating invariant weighted Bergman metrics under biholomorphisms.
method Introducing invariant weight assignments, using Bergman's minimum integral method and domain version of Tian-Yau-Zelditch expansion.
result Uniform convergence of weighted Bergman kernels and metrics on uniform squeezing domains.
Improved covariance matrix estimation for portfolio optimization with guaranteed PSD and controlled conditioning.
problem Guaranteeing positive semidefinite ness and controlling spectral conditioning in IQ estimators.
method Introducing squeezing identity and atomic-IQ parameterization to construct structured channel matrices with PSD guarantees and analytic eigen floor for conditioning control.
result Atomic-IQ improves Sharpe ratios and delivers a more stable risk profile compared to standard estimators.
Tiled Squeeze-and-Excite improves channel attention with local spatial context.
problem Improving channel attention mechanisms in neural networks.
method Proposes tiled squeeze-and-excite (TSE) framework for channel attention.
result Local context of 7 rows or columns is sufficient for matching global context performance.
The Fridman function is bounded by the injectivity radius for certain hyperbolic manifolds.
problem Bounding the Fridman function for hyperbolic manifolds.
method Analyzing the relationship between the Fridman function and the injectivity radius function.
result The Fridman function is bounded above by the injectivity radius function for certain hyperbolic manifolds.
Pinwheels in symplectic rational and ruled surfaces and non-squeezing of rational homology ballsmath.SG
The study explores pinwheels in symplectic surfaces and non-squeezing of rational homology balls.
problem Understanding when Lagrangian pinwheels embed in symplectic rational and ruled surfaces.
method Almost toric fibrations and symplectic rational blow-up.
result A rational homology ball embeds into a rational homology cylinder if and only if the parameter is greater than or equal to 1.
In this paper, we develop improved techniques for defending against adversarial examples at scale. First, we implement the state of the art version of adversarial training at unprecedented scale on ImageNet and investigate whether it remains effective in this setting - an important open scientific question (Athalye et …
Generative classifier derived from any discriminative classifier rejects illegal inputs.
problem Detecting and rejecting illegal inputs like adversarial examples and out-of-distribution samples.
method SDIM-logit: learns generative classifier from logits of any discriminative classifier, imposing statistical constraints.
result SDIM-logit inherits performance of base classifier without loss and can reject illegal inputs.
The generalized coherent states attached to the Jacobi group realize the squeezed states. Imposing hermitian conjugacy to the generators of the Jacobi algebra, we find out the form of the weight function appearing in the scalar product. We show effectively the orthonormality of the base functions with respect to the sc…
Logit dynamics formula reveals self-regulation in softmax policy gradient methods.
problem Understanding the stability and convergence of softmax policy gradient methods.
method Deriving the exact formula for the L2 norm of the logit update vector.
result Logit update magnitudes are modulated by action probability and policy concentration.
Proposes a convex model for mixed logit to handle individual heterogeneity.
problem Non-convex optimization in mixed logit models for individual heterogeneity.
method Sparse and low-rank decomposition for convex formulation.
result Convex formulation avoids simulation-based approximation and unstable model interpretation.
A new method detects and compacts saturated entries in antisparse coding.
problem Efficiently solving antisparse coding problems with -norm penalties.
method Safe squeezing methodology to detect and compact saturated entries, reducing problem dimensionality.
result The method accelerates the computation of antisparse representation by detecting and compacting saturated entries.
MANO normalizes logits to estimate test accuracy without labels.
problem Estimating test accuracy of OOD samples without labels.
method Applies norm to normalized logits.
result Achieves state-of-the-art performance across various architectures.
The study examines Kähler structures on coadjoint orbits of Lie groups using coherent and squeezed states.
problem Does the coadjoint orbits of Lie groups support a Kähler structure?
method Examined three Lie groups: Weyl-Heisenberg, SU(2), and SU(1,1). Used coherent and squeezed states to explore Kähler structures.
result Coherent states provide Kähler embeddings, while squeezed states only symplectic embeddings.
New method uses low logit rank to simplify complex language models.
problem Understanding and learning from modern language models.
method Exploiting the low logit rank structure of language models for efficient learning.
result An efficient algorithm for learning low logit rank models from queries.
Logit models are usually applied when studying individual travel behavior, i.e., to predict travel mode choice and to gain behavioral insights on traveler preferences. Recently, some studies have applied machine learning to model travel mode choice and reported higher out-of-sample predictive accuracy than traditional …
Logit distance bounds representational similarity of models.
problem Approximating linear similarity when distributions are close.
method Defined a logit distance and proved its relationship to representational dissimilarity.
result Logit distance bounds representational similarity, providing nontrivial control in practice.
Proposes a model for clearing prices in financial markets due to margin calls.
problem Determining prices in financial markets following margin calls and short squeezes.
method Developed an explicit formulation for clearing prices after margin calls and short squeezes.
result Identified a threshold short interest ratio leading to discontinuity in clearing prices.
We consider neural network training, in applications in which there are many possible classes, but at test-time, the task is a binary classification task of determining whether the given example belongs to a specific class, where the class of interest can be different each time the classifier is applied. For instance, …