Machine-learning ideas tagged Approximation theory in the Math taxonomy of the Math2NN corpus.
Replace Cox-de Boor evaluation of each cubic B-spline edge activation with its fixed truncated-power expansion. Normalize each scalar edge input to a bounded knot coordinate, evaluate the five shifted cubic positive-part terms in parallel, and contract them with the learned spline coefficients inside one fused kernel.
Build a shallow neural model whose input at every discretization level is embedded into a common Hilbert space with uniformly bounded norm, and constrain every neuron parameter in the corresponding dual norm. The statistical complexity then depends on the Hilbert norm bound rather than the number of retained coordinates, allowing one model design to operate across increasingly fine measurements.
Use the noisy diffusion state to estimate posterior probabilities over data clusters, then route the denoising computation to one or a few cluster-specific score experts. Each expert stores a low-rank covariance or low-rank score correction, so after posterior commitment the denoiser operates at the cluster intrinsic dimension rather than ambient dimension. During the early mixing phase, retain a shared full-dimensional score or evaluate a small weighted set of experts.
Replace an unconstrained scalar MLP certificate with an anchored positive-definite network whose value and gradient are fixed at the equilibrium. Train it so that its Lie derivative along a neural or physical vector field is strictly negative on a prescribed region of attraction. The construction makes stability robust to approximation error: a certificate remains valid whenever the value, gradient, and Lie-derivative errors stay below the target's strict-decrease margin.
Replace the usual sum of pairwise modality similarities with a higher-order score based on the coordinatewise Hadamard product of all normalized modality embeddings. For modalities indexed by i=1,...,m, score a tuple using s(x_1,...,x_m)=\omega^\top(\bar g_1(x_1)\odot\cdots\odot\bar g_m(x_m)), where \omega is learned and \odot is coordinatewise multiplication. This adds explicit m-way interactions without concatenating raw features or introducing a joint encoder.
Replace a full-resolution neural operator with an explicit multiscale reconstruction: a coarse predictor produces a low-resolution field, and a small corrector module combines its gradients and Hessians with learned rapidly varying basis functions. The model should recover fine detail without running the expensive backbone at fine resolution.
Replace a square dense projection in a Transformer or MLP with a trainable recursive butterfly matrix. The layer preserves multiscale channel interactions while constraining every complementary row-column block to rank at most k, reducing parameters and enabling recursive structured matrix-vector products. Unlike an arbitrary sparse layer, the construction has an explicit recursive factorization and a quasi-optimal approximation guarantee among matrices with the same butterfly rank.
Replace a dense neural-network weight tensor with a graph tensor network whose physical modes and internal edge ranks are specified by a sparse rank-adjacency matrix. Unlike tensor-train or hierarchical Tucker layers, the graph can contain selected cycles and skip connections between tensor modes, allowing the factorization topology to match correlations in the weight tensor. Fit the layer with GTN-SVD at a prescribed tolerance and compare accuracy, parameter count, and tensor-contraction…
Replace full-state prediction in a neural simulator or neural operator with prediction of a perturbation around a cheap structured background trajectory. Compute the background defect and known linearized or nonlinear corrections explicitly, and let the neural closure model only the remaining residual. Add a residual-magnitude gate so the learned closure is suppressed when the structured solver already explains the target dynamics.
Replace full-state quantized write-back in a deep low-bit residual stack with quantized increment error feedback. The residual branch quantizes the proposed increment after adding the previous carry, while the carry stores the exact discrepancy; this makes the total error telescope instead of accumulating approximately once per layer.
Group W consecutive diffusion or flow-model loss terms and approximate every intermediate parameter Jacobian by a time-weighted interpolation of the Jacobians at the group’s two endpoints. Sum the intermediate upstream signals into two endpoint cotangents, then perform only two full DiT backward passes instead of W. Add a cosine-similarity gate comparing predicted and actual intermediate velocity changes so that groups violating the local-linearity assumption use exact backpropagation.
Represent every nonnegative equal-mass one-dimensional state by its CDT quantile map relative to a fixed reference density, then train the neural dynamics model in this transformed space rather than on Eulerian grid values. The latent manifold for translations and transport-dominated evolution is substantially flatter: linear transport lies in the span of the initial transformed state and the constant function, while nonlinear conservative dynamics have algebraic approximation error bounds.
Replace an EKF or a large particle ensemble inside a neural world model with a fixed-order polynomial chaos representation of the latent state distribution. The transition network is evaluated under quadrature or sampled chaos variables, and Galerkin projection produces the next uncertainty coefficients directly; a coefficient-wise LMMSE update then assimilates observations without backpropagating through resampling.
Replace a single neural representation of a rapidly oscillatory PDE solution by a macroscopic network plus an explicitly oscillatory corrector network. Feed the network both the slow coordinate $x$ and fast coordinate $y=x/\varepsilon$, and train the resulting composite field in a variational energy objective. This targets the paper's scale-robust approximation bound rather than forcing the optimizer and finite sample set to discover oscillations of wavelength $\varepsilon$.
Replace layer-local reconstruction in post-training quantization with a sequential objective that explicitly cancels the error already accumulated by the quantized prefix. For each layer, quantize its weights so that its local residual approximately negates the propagated incoming deviation, preserving the teacher trajectory even when the codebook is binary or 4-bit.
Replace uniform collocation for a fixed random-feature neural PDE solver with sampling from the leverage-score density of the operator-applied features. Whiten the retained residual feature space before solving for output coefficients, so the sampled least-squares matrix has an identity-like expected Gram rather than inheriting severe anisotropy from the differential operator. The same construction can be used for a linearized neural network by treating Jacobian features as the trial functions.
Replace an autoregressive rollout of a learned dynamical model with a branch-trunk factorization that predicts all future steps simultaneously. The branch network encodes the future action sequence, while the trunk network encodes the current state and query coordinates; their inner products produce the complete horizon. This removes repeated state updates during inference and gives a compact differentiable model for planning.
Represent a one-dimensional sharp-feature signal by a small unordered set of complex singularities and residues instead of predicting all grid amplitudes. A transformer diffusion model predicts these tokens, and a differentiable meromorphic decoder evaluates the result directly at arbitrary coordinates, avoiding grid-specific interpolation and preserving discontinuity structure.
Compress a causal sequence by retaining history positions at equal increments of cumulative representation variation instead of at uniform time intervals. Use the resulting N representatives in a decoder that reconstructs piecewise-constant keys and values; the paper's minimax result predicts a worst-case reconstruction error of total variation divided by 2N, independent of where rapid changes occur.
Replace the standard CFG guidance multiplier in each DDIM step by a coefficient obtained from the terminal guided exponent. Given unconditional and conditional denoiser-derived states D_u and D_c, use r^{1+w}-r instead of w(r-1) on the guidance direction D_u-D_c; this preserves the same two denoiser calls and costs no additional NFE while suppressing low-noise residual blow-up at high guidance.
Attach an online PDE-constrained reconstruction layer to a neural operator instead of accepting its raw output field. The layer stays close to the network prediction at sampled output locations while minimizing a differentiable PDE residual and boundary-condition violation, allowing physics correction for each new input without retraining the neural operator.
Replace the final sequence of diffusion-sampler steps below a positive switching noise scale a with a single analytic normal-mode completion map. Run the existing solver only on [a, sigma_max], then use the denoiser at scale a to extrapolate to the requested terminal floor epsilon. This prevents the step count from growing like log(sigma_max/epsilon) and should preserve the base solver's order when a is coupled to the discretization size.
Compress a trained wide analytic-activation MLP by fitting a narrow same-depth student to the teacher's function values and input derivatives, rather than matching only outputs on a calibration dataset. Choose the student width from the input dimension and target error, with a target scaling m = O((log(1/epsilon))^d_in), and use sequential layer fitting plus channel reweighting to limit error accumulation through depth.
Replace MSE-calibrated scalar quantization with a conservative left-edge quantizer whose thresholds are denser where activation probability and task utility slope are both high. For a monotone utility function, this should preserve high-impact activation regions better than uniform or MSE-optimal bins at the same number of codes, while retaining an explicit rate-versus-quality design rule.