Jumping Connections: Bypassing Locality in Systems

• Jumping Connections are mechanisms that relax strict local interactions by introducing nonlocal transfers across computational and mathematical systems.
• They enable skip pathways in graph neural networks and fixed-distance jumps in quantum walks, improving performance by bypassing traditional layerwise constraints.
• They also manifest in hybrid consensus, wireless sensor routing, and communication complexity, offering strategies to overcome local update limitations.

Searching arXiv for papers on "^^^^^^^^4jumping connections4^^^^^^^^" to ground the article. {"^^^^^^^^4query4 connections4^^^^^^^^\" OR ti:\4^^^^^^^^" connections4^^^^^^^^\"","max_results":^^^^4query4jumping connections4^^^^,"sort_by":"submittedDate","sort_order":"descending"} Searching arXiv for "^^^^^^^^4jumping connections4^^^^^^^^". “Jumping connections” is not a single standardized construction. In the literatures surveyed here, the term and closely related notions denote mechanisms that introduce nonlocal transfer, discontinuous update, or stratified change into systems whose default behavior is local or layer-by-layer. This includes skip pathways in graph neural networks, fixed-distance jumps in discrete-time quantum walks, bundle-type changes on moduli spaces of parabolic connections, jump maps in hybrid consensus and wireless-sensor routing, and pointer-jumping compositions in communication complexity (&&&^^^^4jumping connections4^^^^&&&, &&&^^^^4query4^^^^&&&, &&&^^^^4all:\4^^^^&&&, &&&^^^^4 OR ti:\4^^^^&&&, 0802.2843, Wu et al., 2010).

^^^^4query4^^^^. Taxonomy of the term

A concise way to organize the subject is by the mathematical object on which the “jump” acts. In some settings it is an explicit operator on states; in others it is a structural shortcut between layers; in still others it is a change of stratum or filtration index.

Domain	Meaning of “jumping connection” or cognate notion
Graph neural networks	A skip-connection from an intermediate hidden layer directly to the output
Quantum walks	A fixed-distance permutation that supplements the nearest-neighbor step
Moduli of parabolic connections	A jumping locus where the underlying bundle type changes
Hybrid consensus	A time-driven jump map interleaved with continuous flow
Wireless sensor routing	A jumping transmission mode activated under faults, congestion, or void regions
Communication complexity	Pointer-jumping through layered function composition

A plausible unifying description is that ^^^^^^^^4jumping connections4^^^^^^^^ relax strict locality. In the GCN setting, they bypass depth-induced attenuation; in the quantum-walk setting, they bypass nearest-neighbor transport; in moduli theory, they connect generic and exceptional bundle types; in hybrid systems, they interleave continuous and discrete propagators; and in pointer jumping, they encode traversal through compositions of maps rather than through explicit sequential revelation.

^^^^4all:\4^^^^. Skip pathways in graph neural networks

In the GCN literature represented here, a jumping connection is a skip-connection from the first hidden layer directly to the final output of a two-hidden-layer network. With node features PRESERVED_PLACEHOLDER_^^^^4jumping connections4^^^^, normalized adjacency PRESERVED_PLACEHOLDER_^^^^4query4^^^^, hidden weights PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^ and PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^, output weight $C\in\mathbb R^{k\times m}$, and entry-wise ReLU $\sigma$, the network is

$$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$

The corresponding idealized model replaces $A$ by a sparse effective adjacency $A^*$ that retains only the “essential” edges for propagation (&&&^^^^4jumping connections4^^^^&&&).

The central theoretical result is a joint learning-dynamics and generalization analysis for GCNs with both ^^^^^^^^4jumping connections4^^^^^^^^ and layer-wise graph sparsification. The analysis states that the learned model’s generalization accuracy closely approximates the highest achievable accuracy within a broad class of target functions dependent on $A^*$. The same analysis also identifies an asymmetry between layers: in a two-hidden-layer GCN, generalization is more affected by deviations of the first-layer sparsified matrix from PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^ than by deviations in the second layer. The paper formulates this asymmetry quantitatively through layer-specific conditions such as

PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^

Because PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^, the second layer can be more aggressively sparsified (&&&^^^^4jumping connections4^^^^&&&).

The mechanism is explicit in the node-wise output representation. Writing PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^ and denoting the first-layer skip output by

PRESERVED_PLACEHOLDER_^^^^4query44^^^^

the PRESERVED_PLACEHOLDER_^^^^4query4^^^^5-th node output is

PRESERVED_PLACEHOLDER_^^^^4query4^^^^6

where PRESERVED_PLACEHOLDER_^^^^4query4^^^^7 is the PRESERVED_PLACEHOLDER_^^^^4query4^^^^8-th column of PRESERVED_PLACEHOLDER_^^^^4query4^^^^9. The first term reaches the output without passing through the second hidden layer. This is why first-layer adjacency errors propagate directly, whereas second-layer errors are filtered through the composite term. The paper’s practical guideline is therefore conservative sparsification in the first few hops and more aggressive sparsification in deeper hops, and it states that the same principle extends to deep GCNs with many skip-connections, including Jumping Knowledge networks (&&&^^^^4jumping connections4^^^^&&&).

^^^^4 OR ti:\4^^^^. Fixed-distance jumps in discrete-time quantum walks

In one-dimensional discrete-time quantum walks, jumps are introduced by enlarging the usual walk dynamics on

PRESERVED_PLACEHOLDER_^^^^4all:\4jumping connections4^^^^

The clean walk uses a Hadamard coin

PRESERVED_PLACEHOLDER_^^^^4all:\4query4^^^^

followed by the nearest-neighbor shift

PRESERVED_PLACEHOLDER_^^^^4all:\4all:\4^^^^

so that one clean step is PRESERVED_PLACEHOLDER_^^^^4all:\4 OR ti:\4^^^^. The jump extension adds, at each timestep, a permutation PRESERVED_PLACEHOLDER_^^^^4all:\44^^^^ built from disjoint transpositions of pairs of channels PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^5, where PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^6 is a fixed jump distance. One step of the disordered walk is

PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^7

Two disorder models are distinguished: static disorder, where one permutation is frozen for all times, and dynamic disorder, where the permutation is sampled independently at each timestep (&&&^^^^4query4^^^^&&&).

The asymptotic distributions differ sharply between these regimes. Dynamic disorder yields a Gaussian-like position distribution with linearly growing variance, PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^8. Numerically, for fixed PRESERVED_PLACEHOLDER_^^^^4all:\4^^^^9, the coefficient satisfies PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4jumping connections4^^^^ to leading order, so standard-deviation curves versus PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4query4^^^^ are essentially straight lines. Static disorder instead produces localization and parity-sensitive non-Gaussian profiles. For odd jump size PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4all:\4^^^^, the long-time diagonal probability approaches a stationary profile well-fitted by a discrete Laplace envelope modulated by secondary peaks of period PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4 OR ti:\4^^^^. For even PRESERVED_PLACEHOLDER_^^^^4 OR ti:\44^^^^, the long-time profile develops a three-peak structure: one sharp, localized central peak at PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^5 and two symmetrically displaced broad peaks around PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^6, with small oscillations of period PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^7 superimposed (&&&^^^^4query4^^^^&&&).

A particularly notable numerical result is the universal scaling law for the long-time variance in the static odd-PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^8 case. If

PRESERVED_PLACEHOLDER_^^^^4 OR ti:\4^^^^9

and one rescales

$C\in\mathbb R^{k\times m}$^^^^4jumping connections4^^^^

with empirically determined exponents $C\in\mathbb R^{k\times m}$^^^^4query4^^^^ and $C\in\mathbb R^{k\times m}$^^^^4all:\4^^^^, data for various $C\in\mathbb R^{k\times m}$^^^^4 OR ti:\4^^^^ collapse onto a single U-shaped master curve $C\in\mathbb R^{k\times m}$4, minimal around $C\in\mathbb R^{k\times m}$5. A plausible implication is that the localization width is governed not by $C\in\mathbb R^{k\times m}$6 and $C\in\mathbb R^{k\times m}$7 separately, but by a scaled average jump parameter (&&&^^^^4query4^^^^&&&).

4. Jumping loci in moduli theory and jumping indices in irregular Hodge theory

In algebraic geometry, the language of “jumping” refers not to transport across edges but to changes in geometric type. For rank-$C\in\mathbb R^{k\times m}$8 $C\in\mathbb R^{k\times m}$9-parabolic connections on $\sigma$^^^^4jumping connections4^^^^ with $\sigma$^^^^4query4^^^^, Komyo and Saito stratify the moduli space $\sigma$^^^^4all:\4^^^^ by bundle type

$\sigma$^^^^4 OR ti:\4^^^^

The open stratum $\sigma$4 is the non-jumping locus with $\sigma$5. For $\sigma$6, one has $\sigma$7, and the locus $\sigma$8 is the jumping locus; its preimage in the cyclic moduli $\sigma$9 is an exceptional divisor. Apparent singularities $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$^^^^4jumping connections4^^^^ and dual parameters $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$^^^^4query4^^^^ furnish coordinates on the non-jumping locus and lead to an injective morphism

$$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$^^^^4all:\4^^^^

where $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$^^^^4 OR ti:\4^^^^ is obtained from a blown-up Hirzebruch surface. For $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$4, $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$5 is the blow-up along $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$6, and the map by apparent singularities and dual parameters embeds $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$7 into a further blow-up $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$8 (&&&^^^^4all:\4^^^^&&&).

The geometric content of the jump is explicit in the degeneration process. On a chart with local parameters $$\mathrm{out}(X,A;W,U)=C\,\sigma(WXA)+C\,\sigma(U\,\sigma(WXA)\,A).$$9, one performs an elementary modification of the non-jumping family; as $A$^^^^4jumping connections4^^^^, the underlying bundle changes from $A$^^^^4query4^^^^ to $A$^^^^4all:\4^^^^. The two apparent points coalesce, their duals blow up, and finite symmetric combinations such as

$A$^^^^4 OR ti:\4^^^^

survive as local coordinates on the final blow-up. Here “jumping” denotes passage to a different bundle-type stratum rather than a dynamical hop (&&&^^^^4all:\4^^^^&&&).

A related but distinct use appears in irregular Hodge theory. For an irreducible holonomic connection $A$4 on $A$5 with possibly irregular singularities, Sabbah’s canonical decreasing irregular Hodge filtration

$A$6

has jumps at finitely many real numbers $A$7, called the jumping indices. Jakob and Reiter compute these indices and their multiplicities for four rigid $A$8-connections $A$9. For $A^*$^^^^4jumping connections4^^^^, $A^*$^^^^4query4^^^^, and $A^*$^^^^4all:\4^^^^, all with a single irregular slope $A^*$^^^^4 OR ti:\4^^^^, the nonzero irregular Hodge numbers are

$A^*$4

For $A^*$5, the jumps depend on the relation between parameters $A^*$6 and $A^*$7 and can involve $A^*$8, yielding several distinct multiplicity tables. In this context, a jumping connection is indexed by discontinuities of a filtration rather than by graph-theoretic or dynamical shortcuts (&&&^^^^4query44^^^^&&&).

5. Jump maps in hybrid and communication networks

In multi-agent consensus with time-driven jumps, the system state is hybrid: $A^*$9 evolves continuously and a clock $A^*$^^^^4jumping connections4^^^^ resets at jump times. With flow graph $A^*$^^^^4query4^^^^ and jump graph $A^*$^^^^4all:\4^^^^, Laplacians $A^*$^^^^4 OR ti:\4^^^^ and $A^*$4, and coupling gain $A^*$5, the decentralized dynamics are

$A^*$6

and

$A^*$7

The relevant invariant object is the hybrid multi-consensus subspace

$A^*$8

where $A^*$9 is the coarsest partition of the vertex set that is simultaneously an almost equitable partition for both PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4jumping connections4^^^^ and PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4query4^^^^. Under the bounds

PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4all:\4^^^^

the subspace PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4 OR ti:\4^^^^ is globally asymptotically stable (&&&^^^^4 OR ti:\4^^^^&&&).

In wireless sensor networks, the jump is operational rather than algebraic. The DMRF protocol maintains a Forwarding Candidate Set (FCS) of one-hop neighbors and switches from ordinary hop-by-hop forwarding to jumping transmission mode when a failure, congestion, or void region is detected, or when the remaining-time factor

PRESERVED_PLACEHOLDER_^^^^4query4jumping connections44^^^^

falls below a threshold PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^5. For jumps from node PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^6 to node PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^7, the success ratio and jumping probability are

PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^8

with PRESERVED_PLACEHOLDER_^^^^4query4jumping connections4^^^^9. Failures and acknowledgments update PRESERVED_PLACEHOLDER_^^^^4query4query4jumping connections4^^^^ and PRESERVED_PLACEHOLDER_^^^^4query4query4query4^^^^, and the resulting feedback mechanism shifts probability mass toward reliable neighbors (Wu et al., 2010).

The published simulation setup uses ^^^^4query4jumping connections4jumping connections4^^^^ nodes in a PRESERVED_PLACEHOLDER_^^^^4query4query4all:\4^^^^ area, communication radius PRESERVED_PLACEHOLDER_^^^^4query4query4 OR ti:\4^^^^, MAC/PHY PRESERVED_PLACEHOLDER_^^^^4query4query44^^^^/CSMA-CA at PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^5, and deadlines equal to PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^6. Reported delivery ratios for DMRF are PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^7, PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^8, PRESERVED_PLACEHOLDER_^^^^4query4query4^^^^9, and PRESERVED_PLACEHOLDER_^^^^4query4all:\4jumping connections4^^^^ at faulty-node ratios PRESERVED_PLACEHOLDER_^^^^4query4all:\4query4^^^^, PRESERVED_PLACEHOLDER_^^^^4query4all:\4all:\4^^^^, PRESERVED_PLACEHOLDER_^^^^4query4all:\4 OR ti:\4^^^^, and PRESERVED_PLACEHOLDER_^^^^4query4all:\44^^^^, respectively. Under increasing void-circle radius, average delay remains PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^5, PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^6, and PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^7 for radii PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^8, PRESERVED_PLACEHOLDER_^^^^4query4all:\4^^^^9, and PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4jumping connections4^^^^, while competing hop-by-hop schemes increase sharply (Wu et al., 2010).

6. Pointer jumping and the complexity-theoretic meaning of jumps

In communication complexity, “jumping” refers to evaluation along a layered composition of pointers. In the one-way number-on-the-forehead model with players PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4query4^^^^, the PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4all:\4^^^^-layer pointer-jumping problem has one root in layer PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4 OR ti:\4^^^^, PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\44^^^^ vertices in each intermediate layer, and either PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^5 output vertices for the Boolean problem PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^6 or PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^7 output vertices for the non-Boolean version. The Boolean recursion is

PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^8

and deterministic one-way NOF communication complexity is denoted PRESERVED_PLACEHOLDER_^^^^4query4 OR ti:\4^^^^9 (0802.2843).

The principal upper bound is sublinear for every fixed PRESERVED_PLACEHOLDER_^^^^4query44jumping connections4^^^^:

PRESERVED_PLACEHOLDER_^^^^4query44query4^^^^

In particular,

PRESERVED_PLACEHOLDER_^^^^4query44all:\4^^^^

This rules out an PRESERVED_PLACEHOLDER_^^^^4query44 OR ti:\4^^^^ lower bound for Boolean pointer jumping in the full one-way NOF model for constant PRESERVED_PLACEHOLDER_^^^^4query444^^^^ (0802.2843).

The same paper isolates a restricted class of protocols in which every player is collapsing: player PRESERVED_PLACEHOLDER_^^^^4query44^^^^5 sees the layers behind her individually but the layers ahead only through their composition. Under this restriction, any deterministic one-way NOF protocol for PRESERVED_PLACEHOLDER_^^^^4query44^^^^6 must satisfy

PRESERVED_PLACEHOLDER_^^^^4query44^^^^7

The lower bound is proved using “crossing pairs,” a combinatorial device replacing earlier information-theoretic styles of argument. At the same time, the non-Boolean pointer-jumping problem PRESERVED_PLACEHOLDER_^^^^4query44^^^^8 still admits a collapsing-player upper bound of

PRESERVED_PLACEHOLDER_^^^^4query44^^^^9

when all but one of the layer functions are permutations (0802.2843).

A common misconception would be to treat this usage as analogous to skip-connections in neural networks. The formal resemblance is limited: both involve bypassing naive layerwise progression, but the object of study here is one-way communication under partial information, not representational flow in a learnable architecture.

7. Structural themes and scope

Across these literatures, three motifs recur. First, ^^^^^^^^4jumping connections4^^^^^^^^ alter locality. Fixed-distance permutations in quantum walks supplement nearest-neighbor motion; skip-connections in GCNs bypass depth; DMRF jumps bypass impaired one-hop neighborhoods; pointer jumping composes hidden layers into a single effective map. Second, jumps often induce different asymptotic regimes: Gaussian-like spreading versus localization in quantum walks, layer-sensitive sparsification constraints in GCNs, or constant clusters versus hybrid consensus arcs in multi-agent systems (&&&^^^^4query4^^^^&&&, &&&^^^^4jumping connections4^^^^&&&, &&&^^^^4 OR ti:\4^^^^&&&). Third, in geometry and Hodge theory the jump is stratificational rather than dynamical, marking discontinuous changes of bundle type or filtration grade (&&&^^^^4all:\4^^^^&&&, &&&^^^^4query44^^^^&&&).

This suggests that “^^^^^^^^4jumping connections4^^^^^^^^” functions best as a family resemblance term rather than a universal definition. Its precise content depends on whether the ambient structure is a Hilbert-space evolution, a neural message-passing architecture, a moduli stack, a hybrid dynamical system, a routing protocol, or a communication game. What remains stable across these cases is the introduction of a mechanism that bypasses an underlying local rule while preserving enough structure to admit rigorous analysis.