PolarBM: Complex-valued Boltzmann Machine for Modeling Audio Signals in Polar and Log-polar Coordinates

2026-07-14 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Audio & Speech Processing · Depth: Expert, quick

Summary

A novel Boltzmann machine, PolarBM, is proposed to naturally model complex-valued data, such as audio signal spectra, in polar coordinates. Unlike conventional real-valued machine learning methods that discard amplitude-phase relationships, PolarBM defines a probability density function where phase explicitly depends on amplitude. An extension, LogPolarBM, further models amplitude on a logarithmic scale to align with human auditory perception, utilizing a power-weighted noncentral complex Gaussian (PW-NCCG) distribution. Practical restricted variants, PolarRBM and LogPolarRBM, are also introduced. Experimental results, published on 2026-07-14, demonstrate that these proposed RBMs achieve superior modeling accuracy compared to deep neural networks by explicitly capturing the amplitude-phase dependency. While initially focused on audio, the models' utility extends to diverse fields like wireless communications and quantum mechanics.

Key takeaway

For Machine Learning Engineers developing models for complex-valued data, particularly audio signals, you should consider PolarBM or LogPolarBM. These models offer superior accuracy by explicitly capturing the inherent amplitude-phase dependency, a structural information often discarded by real-valued frameworks. Implementing these Boltzmann machine variants can lead to more accurate representations and better performance in applications ranging from audio processing to wireless communications, providing a significant advantage over conventional deep neural networks.

Key insights

Modeling complex-valued data in polar coordinates, explicitly linking amplitude and phase, improves accuracy.

Principles

• Complex data benefits from explicit amplitude-phase modeling.
• Logarithmic amplitude scaling aligns with human perception.
• Boltzmann machines can be adapted for complex variables.

Method

PolarBM defines a probability density for complex variables where phase depends on amplitude. LogPolarBM uses a PW-NCCG distribution for logarithmic amplitude modeling. Restricted variants (RBMs) are derived.

In practice

• Apply to audio signal processing.
• Explore in wireless communications.
• Investigate for quantum mechanics.

Topics

• Complex-valued Data
• Boltzmann Machines
• Audio Signal Processing
• Amplitude-Phase Modeling
• Log-polar Coordinates
• Machine Learning Models

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Calibrated Harmonic Overlaid Implicit Neural Representations for Multi-Dimensional Data — CHOIR enhances Implicit Neural Representations by using harmonic superposition and spectrum calibration for stable, deeper networks and improved data recovery.
• ZEBRA: Zero-Shot Entropy-Regularized Prompt Learning for Base-to-Novel Generalization in Audio-Language Models — Prompt learning in ALMs creates a base-to-novel generalization gap; ZEBRA mitigates this by fusing logits and entropy regularization.
• Physics-Aware Complex-Valued State Space Model with Scattering-Prior Feature Modulation for PolSAR Image Classification — CV-SSMNet integrates physics-aware complex-valued state-space models and scattering-prior modulation for enhanced PolSAR image classification.
• IQ-JEPA: A Joint-Embedding Predictive Architecture with a Hermitian Vision Transformer for Sound Speed and Attenuation Estimation from Ultrasound IQ Data — IQ-JEPA uses a Hermitian Vision Transformer and self-supervision to efficiently estimate ultrasound sound speed and attenuation from IQ data.
• Quality Adaptive Angular Margin Learning for Respiratory Sound Classification — QLung uses quality-adaptive angular margins to improve respiratory sound classification generalization and handle class imbalance.
• An explicit operator explains end-to-end computation in the modern neural networks used for sequence and language modeling — SSMs compute by encoding inputs as traveling waves in a nonlinear oscillator network.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.