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 Muon's single momentum matrix with a weighted mixture of fast and slow relaxation modes. The fast mode tracks rapidly changing gradients while the slow mode preserves a longer-horizon direction; their mixture is semi-orthogonalized and applied as the matrix update.
Replace an unconstrained neural ODE or recurrent update field with the negative gradient of a learned scalar energy \(E_\theta(z,t)\). The resulting hidden-state dynamics have an exact Lyapunov certificate: energy decreases continuously, bounded trajectories cannot exhibit nonstationary recurrence, and the Łojasiewicz mechanism predicts convergence to a single equilibrium rather than persistent oscillation or chaos.
Attach a recursive Bayesian state estimator to a neural sequence classifier. The network produces per-step emission likelihoods, while a persistent Markov transition model propagates beliefs between steps; when inputs are missing, marginalize the missing emission instead of replacing it with a sentinel or arbitrary imputation. This should suppress isolated logit oscillations and remain robust when missing data arrive in bursts.
Select model-based rollout branches using a task-gated log-determinant information objective, so the planner receives counterfactuals that are both decision-relevant and nonredundant. Add a conflict-projection step that removes branches whose predicted actions or outcomes disagree with the trusted policy in an unsafe or credibility-sensitive way, then validate a fixed batch before policy updates.
Freeze a wide neural spatial dictionary, then compress and whiten it using the quadrature mass matrix before solving for output coefficients or latent PDE states. The retained basis removes feature directions that are numerically invisible or nearly dependent under the actual domain discretization, while preserving the represented function space up to the chosen SVD rank.
Constrain a recurrent neural transition to map a compact learned-state region strictly into its interior, creating a neural analogue of the paper's maximal attractor. Unlike simple spectral normalization, this permits a nontrivial invariant set and can preserve task-relevant recurrent dynamics while preventing long-horizon state escape.
Constrain a recurrent interaction matrix to a directed cycle, or initialize it near a cyclic block structure, and certify that the intended unstable or oscillatory mode survives independent gain perturbations. The cyclic topology makes the full network characteristic equation exactly reducible to one scalar loop equation. A robustness penalty can then preserve long-horizon oscillations under quantization, dropout-like gain errors, pruning, or hardware variation.
Replace simultaneous parameter updates with sequential block updates whose order is selected using estimated cross-block sensitivity. The paper shows that sequential policy updates can have a substantially smaller local contraction factor than decoupled or differently ordered updates; the neural analogue is to order attention, normalization, backbone, and head blocks according to the spectral radius of their composed update map.
Add sparse directed coupling between parallel neural modules, recurrent states, or distributed replicas so that each module is driven toward a common trajectory without forcing an undirected or balanced communication graph. Select n-1 directed paths per strongly connected component and assign gains using the estimated Lipschitz bound of the uncoupled module; activate the coupling only when its graph-certified strength exceeds the predicted synchronization threshold.
Replace pointwise sequence reconstruction with reconstruction of overlapping past and future Hankel windows in a shared latent manifold. A first encoder compresses the delay-coordinate trajectory, while a second decoder or predictor reconstructs the future block from the latent state; training therefore penalizes representations that fit observations but do not preserve dynamical evolution.
Certify during or after RL training that a neural policy keeps the closed-loop state inside a prescribed safe set under bounded disturbances and observation errors. Use spectral normalization or a Lipschitz penalty to reduce policy gain, then compute a conservative one-step safety margin that must remain positive over reachable states.
Replace signed-input aggregation in a spiking recurrent cell with a causal micro-event queue that processes excitatory and inhibitory arrivals in timestamp order, applying threshold and reset after each event. This preserves computations that disappear when all events in a timestep are replaced by one net current, particularly near threshold.
Wrap a neural policy or sequence-model controller with an online-estimated ultra-local model of a scalar safety output, such as distance-to-obstacle, queue length, battery margin, or constraint slack. Estimate the unknown drift and control effectiveness directly from recent observations, then impose a robust control-barrier constraint that subtracts an empirical uncertainty envelope before allowing the neural action.
Use a projected dual variable as a feedback controller for terminal feasibility rather than selecting a fixed penalty coefficient. The multiplier increases after infeasible batches and decreases after feasible batches, with an explicit cap and drift-balance diagnostic that detects whether the policy-dual loop is stable.
Replace a learned critic with group-relative trajectory advantages whose weights are explicitly ordered by terminal feasibility. Feasible rollouts receive larger positive update weight than violating rollouts, while per-timestep normalization prevents high-variance late-horizon returns from dominating the policy gradient.
Use the estimated distance to a saddle-node ghost as an inference-time controller for recurrent or neural-ODE computation. Far from a fold, take large integration steps or update only the fast state; near the fold, reduce the step size or allocate extra recurrent evaluations because the state is expected to linger and become sensitive to small parameter changes.
Augment each recurrent or state-space hidden channel with a two-dimensional oscillatory state and periodically compute a pseudo-phase from its Cartesian coordinates. Use sparse event-triggered feedback to reduce the squared phase order parameter, preventing hidden channels from synchronising while avoiding the computation and communication cost of continuously recomputing the control signal. The controller acts as a tangent rotation of each two-dimensional hidden state, changing phase diversity…
Turn the latent substrate into a persistent computational workspace for sequential inputs: each new observation is written into a designated subspace, processed by the same local rule, decoded, and then selectively retained or reset. This creates a compact recurrent model whose state can accumulate algorithmic information across a stream without expanding the parameter count.
Use the paper's structure-exploiting primal-dual active-set strategy to solve barrier-constrained neural updates without invoking a generic quadratic-program solver at every step. The active constraints identify which layers or state statistics are actually close to instability, while warm-started multipliers and active sets should make the safety correction nearly constant-cost when the training trajectory changes smoothly.
Replace a single residual stream or unconstrained hyper-connection with S parallel feature streams whose cross-stream mixing matrix is doubly stochastic. Parameterize the matrix with Sinkhorn normalization so every layer preserves total stream mass while still learning adaptive information routing. This is a low-overhead alternative to dense cross-stream attention and should reduce stream explosion, collapse, and sensitivity to depth.
Treat a neural-network training run as a time-dependent dynamical system and define scalar late-time features that distinguish convergent, oscillatory, noisy, and divergent regimes. Instead of exhaustively sweeping a two-dimensional hyperparameter grid, continue the threshold curve of a feature in the learning-rate/weight-decay or learning-rate/noise plane using a secant predictor and one-dimensional correction sweep. This produces an automatically updated stability map and can be used to keep…
Monitor learning as the ratio of future-task value gained to information irreversibly acquired by an update, rather than treating every reduction in training loss as equally productive. Penalize updates that absorb substantial data-specific information without increasing deletion-counterfactual value, and use the ratio to stop, trust-region, or schedule updates. This creates a falsifiable diagnostic for overfitting without assuming that overfitting and low efficiency are monotonically related.
Use the theta-SRG of each residual-block Jacobian to regularize its gain and phase spread, rather than constraining only its spectral norm. For an implicit or deeply unrolled residual network, maintain a positive distance between the SRG enclosure of the block composition and the critical feedback point -1, giving a directly testable invertibility margin for long-horizon propagation.
Treat a change in a neural network mask, expert set, layer width, or adapter configuration as an optimal transition problem rather than an instantaneous switch. A cheap planner proposes a short sequence of topology masks and parameter interpolations, while an expensive forward-pass feasibility filter rejects each candidate intermediate model if it violates accuracy, activation, norm, latency, or memory limits. This permits dynamic pruning and MoE reconfiguration with a certificate that the…