Every idea extracted from recent arXiv mathematics papers — verified and unverified. Click an idea to open its full card; badges show the empirical verdict.
Replace dense attention over structured object tokens with attention over role-filler tensor-product representations. A learned query specifies both a role and a filler, retrieves objects matching that binding, extracts a target role, and rebinds the extracted filler into an output object.
Replace the global EMA update for each linear-layer momentum matrix with a delta-rule update that learns the current output-side gradient value only along the current input-key direction. Frequently occurring directions are corrected repeatedly, while rarely visited directions are not unnecessarily overwritten or uniformly decayed. Use the resulting matrix as the ordinary momentum buffer in SGD, AdamW, or another optimizer.
Wrap an observation-based neural policy with a real-time safety filter that accounts for uncertainty in its latent-state estimate. The policy proposes an action, while a quadratic program minimally modifies that action so a control-barrier/value function remains nonnegative for every state inside a conformally calibrated error set.
Use an online estimate of the positive feedback gain among logits, routing probabilities, and representations to adjust the softmax temperature. Increase temperature when the estimated cyclic gain approaches the instability regime, preventing exponential amplification and router collapse without globally weakening all layers.
Replace the usual random or elementwise-positive linear-attention feature map with a rank-one positive-semidefinite feature map derived from query and key vectors. For normalized inputs, the resulting kernel is the squared inner product, which is nonnegative and gives a geometrically structured interference pattern that is better suited to associative recall than an arbitrary low-rank feature map.
Replace unconstrained residual updates with blocks whose Jacobian is monitored through a Davis–Wielandt shell. The shell simultaneously measures directional dissipation and non-normal amplification, yielding a per-block step-size or residual-scale bound that is stronger than checking only the largest eigenvalue or spectral norm.
Compress each hidden layer by retaining directions that are simultaneously reachable from the observed input distribution and observable at the network output. Unlike PCA or SVD, the retained subspace is weighted by downstream task sensitivity, so high-variance but output-irrelevant directions can be removed while low-variance predictive directions are preserved.
Replace an unconstrained token mixer or feed-forward residual map with a feature-conditioned rational operator whose transfer matrix is positive real on the positive orthant. Apply it through a resolvent, rather than an additive residual, so that the accretivity certificate yields a nonexpansive implicit update and suppresses activation explosions.
Replace ordinary row-softmax attention with a doubly stochastic Sinkhorn attention plan W, and periodically recover a gauge-fixed pairwise cost from W using the exact double-centering identity. Use this recovered cost to initialize or regularize a structured attention score, making the attention geometry identifiable despite arbitrary query and key row and column offsets.
Augment a 1D neural operator or transformer with explicit tokens for detected discontinuities. Advance each front analytically using the local Rankine–Hugoniot speed and train the network only to reconstruct smooth regions and the residual caused by source terms and grid resolution.
Insert a fixed DPSS/prolate projection before an expensive neural block, retaining exactly the modes whose time-frequency concentration eigenvalues exceed a target threshold. Use the paper's tail-quantile formula to choose the projection rank from sequence length, effective bandwidth, and tolerated energy loss, then optionally learn a small correction in the retained coordinates. Unlike a Fourier truncation, the basis is optimized for simultaneous localization in the finite input window and the…
Replace a conventional whitening transform with a constrained whitening layer that minimizes cross-channel covariance while requiring every output channel to remain correlated with its designated input channel by at least a threshold \(\rho_{\min}\). The layer exploits the orthogonal freedom in whitening to find a rotation that preserves channel identity instead of arbitrarily mixing features. It can be inserted before an MLP, convolution, or attention projection and compared directly against…
Insert a linear Johnson–Lindenstrauss bottleneck around a set of jointly processed representations, choosing its width from the sharp finite-set dimension bound rather than from the model's nominal hidden size. The projection should preserve pairwise distances between tokens, patches, or retrieved items, allowing a downstream attention or MLP block to operate at lower width while retaining the geometry relevant to similarity computations.
Add an explicit cumulative damage state to a neural sequence model and penalize predictions whose degradation estimate decreases as this state increases. This transfers the paper's separation of physics-informed history encoding and monotonicity regularization to battery-health prediction, remaining-useful-life estimation, thermal aging, and other nonstationary sequence problems.
Use predicted covariance reduction as a differentiable gate for selecting tokens, views, sensors, or retrieved demonstrations. The gate favors inputs with high expected information gain while accounting for acquisition cost, turning attention and data collection into active observability optimization.
Replace the usual momentum schedule in a neural-network optimizer with a discretization of the paper's lemniscate-acceleration ODE. The method uses a time-dependent friction coefficient that is initially very large and then decays according to lemniscate sine and cosine functions, targeting faster reduction of the gradient norm than constant-momentum SGD or standard Nesterov schedules.
Use the evolving singular spectrum of the represented matrix W_t=U_tV_t^{\top} to modulate one common, gauge-equivariant learning rate. Slow the shared update when spectral mass accumulates outside the intended low-rank subspace, preventing adaptive dynamics from amplifying nuisance tail directions while retaining the shared-rate structure needed for low-rank recovery.
Treat undesirable neural-network states as obstacles and steer training or inference away from them with a smooth distance barrier while preserving a nominal loss descent direction. The barrier can protect against exploding activations, excessive attention concentration, unsafe controller outputs, or leaving a certified representation region without introducing discontinuous gradient clipping.
Train masked predictors with an explicit mixture of high-visibility masks, low-visibility masks, and a small atom at the fully masked input. High-visibility masks preserve ordinary denoising quality, while low-visibility and fully masked examples force the network to learn global mode frequencies that are invisible when nearly all context is shown. Tune the low-visibility mass using unconditional-mode recovery as an auxiliary validation metric.
Replace a static batch size with a schedule optimized for a prescribed learning-rate schedule and a fixed total number of processed examples. Steps whose stochastic-gradient noise has a large effect on the paper's loss bound receive larger batches, with the weighting determined by the future learning-rate tail rather than by a hand-designed warmup or cooldown rule.
Encode a continuous or sampled multivariate time series as degree-N log-signatures on m local intervals, choosing (m,N) to minimize the number of stored scalars under an explicit linear-CDE approximation budget. Unlike fixed-size temporal patches, the representation increases temporal resolution when the path is oscillatory and increases algebraic degree when the path is smooth but strongly nonlinear.
Train a cheap shared multi-task probe briefly, extract one semantic embedding per task, and use density-based clustering to determine which tasks should share a neural trunk. After clustering, replace the globally shared trunk by one trunk per discovered cluster, with task heads remaining separate; this preserves cooperation among related tasks while isolating destructive task interactions.
Replace deterministic LoRA importance scores with one-sided tests of whether each rank-one update has population contribution at least a user-selected threshold. Maintain empirical contribution samples during fine-tuning, estimate their uncertainty, and prune the components with the weakest statistical evidence while respecting the target rank budget. The method should avoid deleting components merely because their latest minibatch gradient was small.
Construct a robust covariance estimate of layer activations by replacing each feature with its empirical Gaussian normal score before eigendecomposition, then applying coordinate-wise nonlinear eigenvalue shrinkage rather than multiplying all eigenvalues by one scalar. Use the cleaned covariance to whiten activations or precondition updates to the associated linear layer. This targets unstable directions caused by small batches, heavy-tailed activations, and rare outliers while retaining…