When do complex-valued neural networks help? A study of representation, geometry, and optimization

• URL: http://arxiv.org/abs/2605.27673v1
• Date: Tue, 26 May 2026 20:49:23 GMT
• Title: When do complex-valued neural networks help? A study of representation, geometry, and optimization
• Abstract summary: We show that complex-valued inputs alone do not determine when complex arithmetic improves learning.<n>Across synthetic RF tasks, complex representations are useful but not universally superior.<n>CVNNs are best viewed as structured inductive biases whose gains depend on representation, symmetry, and optimization.
• Score: 1.4935265696074094
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Complex-valued Neural Networks (CVNNs) are often motivated by domains where information is naturally encoded in magnitude and phase. Yet complex-valued inputs alone do not determine when complex arithmetic improves learning: the label signal may lie in amplitude, phase, their coupling, or a symmetry that real-valued models can also represent under suitable coordinates. We study this through a representation-first evaluation of CVNNs against Cartesian real, polar, phase-only, magnitude-only, parameter-matched real, and FLOP-matched real baselines. Across synthetic RF tasks, complex representations are useful but not universally superior. PSK-only tasks favor phase-aware and complex-valued models, QAM-only tasks favor magnitude-based models, mixed PSK+QAM gives only a small complex-valued advantage, and unseen carrier-phase rotations break coordinate-dependent models without augmentation. Similar patterns appear beyond RF: in quantum-wavefunction prediction, momentum is invisible to $|ψ|$ but recoverable from phase, while EEG analytic-signal experiments show that phase locking, amplitude bursts, and phase-amplitude coupling each favor different coordinate views. We also identify a benchmarking artifact on RadioML 2018.01A. Under matched-shared-trial selection, a CReLU complex model exceeds the best real baseline by 22.94 PP; under independent per-family tuning on the same data and 16-trial search space, the gap collapses to 2.46 PP. Gradient analysis traces the inflated gap to high-learning-rate first-step instability in real baselines, while complex parameter coupling distributes the loss signal more robustly. A learning-rate $\times$ activation factorial confirms the failure is primarily hyperparameter-driven. Overall, CVNNs are best viewed as structured inductive biases whose gains depend on representation, symmetry, and optimization, not as universally superior architectures.

Related papers

• On a joint simultaneous learning of relevant feature subsets and subspaces in regression-like problems [1.0152838128195467]
  We coin as Entropy-optimal Manifold Regression (EOMR)<n> EOMR is compared to the most complete set of state-of-the-art tools from the Artificial Intelligence (AI) and Machine Learning (ML)<n>For a Hasegawa-Wakatani example, EOMR distills a very simple entropy-optimal and skilful description of the leading Essential Orthogonal Function (EOF) dynamics.
  arXiv  Detail & Related papers  (2026-07-30T11:51:29Z)
• Modeling Normal Is All You Need: Joint Latent Clustering for Anomaly Detection in Multimodal Cyber-Physical Systems [47.027290803102666]
  Normal behaviour is the union of many imbalanced, curved, thin-fringed operating regimes.<n>We model the normal law with a jointly learned latent representation plus explicit Gaussian-mixture mode clustering.<n>We evaluate under a deliberately fair protocol: raw point-wise metrics with no point adjustment, a trivial-detector difficulty split, prevalence-matched F1, and train-normal-only calibration.
  arXiv  Detail & Related papers  (2026-07-07T10:10:00Z)
• Effective Covariance Dynamics in Solvable High-Dimensional GANs [2.639628765667673]
  We study a solvable high-dimensional model of generative adversarial network (GAN) training in which a linear generator learns a low-dimensional subspace from data.<n>We prove that the microscopic training process converges, in the high-dimensional limit, to deterministic ordinary differential equations governed by this effective covariance.
  arXiv  Detail & Related papers  (2026-06-25T16:33:54Z)
• Influence-Guided Symbolic Regression: Scientific Discovery via LLM-Driven Equation Search with Granular Feedback [56.69850045068714]
  We introduce textitInfluence-Guided Regression (IGSR), a method that frames equation discovery as an iterative two-step process.<n>We demonstrate IGSRs effectiveness on a diverse suite of benchmarks, including LLM-SRBench, pharmacological PKPD models, an epidemiological simulation, and real-world genomic data.
  arXiv  Detail & Related papers  (2026-05-27T23:48:01Z)
• Deterministic Mapping of Topological Phases via Autoregressive Exogenous Neural Networks [0.0]
  Three dynamic neural network (NN) architectures -- NAR, NARX, and NIO -- are evaluated.<n>We show that NARX achieves superior predictive fidelity, reaching a Mean Squared Error (MSE) of $10-27$.<n>We observe a "complexity paradox" where the NARX model's accuracy collapses at higher delays.
  arXiv  Detail & Related papers  (2026-05-26T17:11:22Z)
• Efficient Adjoint Matching for Fine-tuning Diffusion Models [47.05085426361986]
  We propose EAM, which substantially improves training efficiency by reformulating the SOC problem with a textit base drift and a textitterminal cost.<n>EAM converges up to 4x faster than AM and matches or surpasses it across various metrics including PickScore, ImageReward, HPSv2.1, CLIPScore and Aesthetics.
  arXiv  Detail & Related papers  (2026-05-12T03:55:12Z)
• Neural Precoding in Complex Projective Spaces [39.36151459385294]
  Deep-learning models are trained to map features derived from channel coefficients to labels derived from precoding weights.<n>Traditionally, complex-valued channel and precoder coefficients are parameterized using either their real and imaginary components or their amplitude and phase.<n>We propose a framework based on complex projective space (CPS) parameterizations of both the wireless channel and the weighted minimum mean squared error (WMMSE) precoder vectors.
  arXiv  Detail & Related papers  (2026-03-08T21:29:26Z)
• Is Flow Matching Just Trajectory Replay for Sequential Data? [46.770624059457724]
  Flow matching (FM) is increasingly used for time-series generation.<n>It is not well understood whether it learns a general dynamical structure or simply performs an effective "trajectory replay"<n>We show that the implied sampler is an ODE whose dynamics constitutes a nonparametric, memory-augmented continuous-time dynamical system.
  arXiv  Detail & Related papers  (2026-02-09T06:48:45Z)
• Timely Parameter Updating in Over-the-Air Federated Learning [45.5660377179285]
  We propose Freshness Freshness-mAgnItude awaRe top-k (FAIR-k), an algorithm that selects, in each communication round, the most impactful subset of gradients to be updated over the air.<n>We show that FAIR-k promotes fresh (and fair) parameter updates and enhances communication efficiency by enabling an extended period of local training without significantly affecting overall training efficiency.
  arXiv  Detail & Related papers  (2025-12-22T07:18:13Z)
• Towards Generalized Synapse Detection Across Invertebrate Species [0.07999703756441755]
  SimpSyn is a single-stage Residual U-Net trained to predict dual-channel spherical masks around pre- and post-synaptic sites.<n>It consistently outperforms Synful in F1-score across all volumes for synaptic site detection.
  arXiv  Detail & Related papers  (2025-09-21T11:40:49Z)
• Equivariant U-Shaped Neural Operators for the Cahn-Hilliard Phase-Field Model [4.79907962230318]
  We show that an equivariant U-shaped neural operator (E-UNO) can learn the evolution of the phase-field variable from short histories of past dynamics.<n>By encoding symmetry and scale hierarchy, the model generalizes better, requires less training data, and yields physically consistent dynamics.
  arXiv  Detail & Related papers  (2025-09-01T09:25:31Z)
• Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression [53.15502562048627]
  Recent work has built the connection between self-supervised learning and the approximation of the top eigenspace of a graph Laplacian operator. This work delves into a statistical analysis of augmentation-based pretraining.
  arXiv  Detail & Related papers  (2023-06-01T15:18:55Z)
• Autoencoder-driven Spiral Representation Learning for Gravitational Wave Surrogate Modelling [47.081318079190595]
  We investigate the existence of underlying structures in the empirical coefficients using autoencoders. We design a spiral module with learnable parameters, that is used as the first layer in a neural network, which learns to map the input space to the coefficients. The spiral module is evaluated on multiple neural network architectures and consistently achieves better speed-accuracy trade-off than baseline models.
  arXiv  Detail & Related papers  (2021-07-09T09:03:08Z)
• Kernel and Rich Regimes in Overparametrized Models [69.40899443842443]
  We show that gradient descent on overparametrized multilayer networks can induce rich implicit biases that are not RKHS norms. We also demonstrate this transition empirically for more complex matrix factorization models and multilayer non-linear networks.
  arXiv  Detail & Related papers  (2020-02-20T15:43:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.