Attention variants, state-space and recurrent cells, normalization and token-mixing schemes — each tested against the standard block it replaces.
Replace a single smooth neural vector field with a finite collection of smooth subnetworks selected by learned affine hyperplanes. The architecture exposes switching geometry directly, allowing it to represent friction-like or threshold dynamics without approximating discontinuities using excessively steep activations.
Replace a fixed-size learned latent for an array-valued complex tensor with a variable-length list of continuous rank-one spectral atoms. An encoder predicts candidate receive direction, transmit direction, residual off-grid offsets, and complex gains; the decoder reconstructs the tensor analytically from the array-response formula, so changing the antenna dimensions does not require changing the decoder weights.
Replace fixed random Fourier features in a coordinate network or PINN with a small complex Fourier dictionary whose propagation directions are learned from a weighted residual. Given directions, solve the linear feature coefficients exactly or by ridge regression, and optimize only the directions in the outer loop. This should represent low-directional-complexity fields with fewer features and avoid wasting gradient updates on coefficients that can be fitted analytically.
Attach a deterministic supervisory automaton to a neural policy or sequence model and mask every event disabled by the current supervisor state. Use a short receding-horizon planner over admissible events to resolve conflicts between neural preferences and shared-resource constraints. The network scores useful actions, while the automaton supplies an exact safety layer.
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…
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.
Replace a stack of independently parameterized residual or MLP blocks with a small latent grid or vector repeatedly updated by one shared transition rule. Let the number of updates depend on the current latent state, so easy examples terminate early while hard examples receive more computation, potentially improving parameter efficiency and extrapolation.
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.
Use trajectory sensitivities to remove neural units or parameter groups whose effects are redundant over the available data support. A parameter group is pruned when its Fisher contribution is small or its sensitivity is nearly collinear with other groups, producing a compact neural ODE without relying only on parameter magnitude.
Replace blind width or rank expansion with a residualized candidate-selection procedure for low-rank adapters or expert banks. Each candidate block is measured after removing the component already represented by the incumbent network, filtered for numerical conditioning, and accepted only when its projected response produces sufficient loss-relevant gain. This should produce smaller adaptive networks than adding candidates by fixed order or magnitude heuristics.
Replace an oversized recurrent hidden state or raw history stack with a causal filtered input-output lift followed by an SVD-selected bottleneck. The actor, critic, and Bellman regression operate only on the identifiable memory coordinates, preventing deterministic null directions from being fitted as if they were independent state variables.
Attach an evidential cost head to a neural graph model, representing each edge cost by a weighted set of interval boxes, and compress this representation before the downstream shortest-path or routing solver. Instead of minimizing Jaccard or Jousselme distance between the original and compressed mass functions, choose merges that minimize the induced cost error on the currently selected route, while enforcing a conservative monotonicity condition so that the resulting path regret is bounded.
Replace independent token-to-expert softmax routing with a fixed-budget congestion game. Each token group distributes a fixed routing mass across experts, while the marginal value of an expert decreases as other groups send mass there. Iteratively route toward the highest current marginal utility and exploit sorted-prefix supports to produce sparse, capacity-aware assignments.
Add an online receptive-field expansion monitor to a graph neural network and use it to gate message-passing depth or invoke graph pooling. For a sampled node set F and propagation neighborhood K, continue fine-scale propagation only while the growth ratio |KF|/|F| is close to one; when it is persistently expansive, replace further propagation with pooling, local attention, or long-range skip messages. This transfers the paper's Følner-versus-paradoxical mechanism into an architecture-level…
Replace dense continuous action search during neural-controller verification with a finite set of representative inputs induced by affine pieces of the interval neural dynamics. This makes safety checking parallel over state cells and candidate actions, enabling much cheaper certification or repeated safe-set updates.
Replace ordinary sum or mean pooling in a permutation-invariant set network by the complete family of basic tropical multisymmetric values. For an input set of n points in R^r, each feature computes the maximum total coordinate score obtainable by assigning disjoint rows to prescribed coordinate channels. The resulting representation is invariant to row permutations, separates all multisets, and inherits a bi-Lipschitz relation to optimal row matching, so nearby sets cannot be arbitrarily…
Replace the single global MoE capacity factor with expert-specific capacity reservations chosen from a small reliability menu. Experts with highly variable or operationally important token loads receive larger robust buffers, while predictable experts run closer to their mean load. This should reduce token dropping and padding waste simultaneously, especially under distribution shift or bursty routing.
Replace sequential upward message passing in a tree-structured neural module with rake–compress contraction of quadratic latent-state messages. Each node stores a quadratic value function and each edge stores a linear transition or coupling triple; leaf elimination and unary-node compression are implemented as batched Schur complements, followed by a reverse pass that reconstructs node latents and edge outputs. The layer is exact for Gaussian or quadratic latent models and remains…
Represent a hybrid trajectory with one neural module per known dynamical phase rather than a single network spanning all phases. Feed the predicted terminal state of phase r directly as the initial state of phase r+1, so continuity is satisfied by construction instead of by a soft interface penalty. This should improve learning near abrupt changes and remove an otherwise poorly conditioned loss-weight tradeoff.
Replace an unconstrained image denoiser or refinement block by a gradient step on an input-convex neural potential. The resulting map has a verifiable nonexpansiveness guarantee when the potential is convex and its gradient is sufficiently smooth, reducing error amplification across repeated applications and making the module safer under distribution shift.
Replace the shared hidden trunk of a multi-output regression network with a small bank of differentiable symbolic units, then let every output use a sparse additive or multiplicative combination of the same units. The architecture explicitly tests whether outputs share a latent mechanism instead of merely sharing arbitrary neural features, improving identifiability and producing equations that can be inspected or exported.
For each frozen weight tensor, append a second tensor of identically shaped zero weights and assign trainable scores to both the real and dummy edges. Select a fixed number of candidates by top-k score in the doubled space; real edges selected by the competition remain active, while selected dummy edges consume the quota without changing the network. The resulting number of active original edges is learned rather than imposed by a separate layerwise sparsity search.
Parameterize a trainable weight update as \(\Delta W=UV^{\top}\) with an excessive initial rank \(r\), and penalize active columns using an exact column \(\ell_{2,0}\) penalty. Increase \(\lambda\) along a warm-started path and hard-delete redundant paired columns, producing an automatically selected rank without training a separate model for every candidate rank. Apply scale balancing after each update so pruning decisions are invariant to reciprocal rescaling of factor pairs.
Let a neural policy emit an unconstrained abstract action z, then solve a state-dependent feasibility problem that maps z to an admissible optimal-control parameter p before execution. Unlike coordinate-wise clipping, the mapping accounts for predicted dynamics, coupled state and input constraints, and recursive feasibility, allowing the policy to retain a simple unconstrained output space while the controller enforces plant constraints.