Residual additive networks improve speech recognition with exact data routing
Task-Directed Residual AddUNet:Perfect-Reconstruction Routing for Full-Rate Representations
Machine Learning
Summary
High-quality machine learning often needs ways to break down and combine data exactly without losing information. The paper shows that a special kind of neural network called AddUNet can perfectly reconstruct its input even as it filters and routes parts of data for tasks like speech recognition. The authors introduce a more flexible version that keeps all data perfectly intact while learning to send irrelevant parts away from the main processing path. This helps improve speech recognition accuracy without sacrificing data quality or needing extra reconstruction steps.
What this means in practice
- •For speech recognition engineers: Improve speech recognizer front-end by precisely routing irrelevant signals while preserving critical information to reduce error rates.
- •For machine learning system builders: Design neural network modules that ensure exact data reconstruction while selectively routing features relevant to specific tasks.
Tested on one dataset.
Authors
Vikram R. Lakkavalli
Abstract
This paper establishes a perfect-reconstruction (PR) interpretation of AddUNet and its full-rate realization, and introduces a Residual Full-Rate PR architecture for task-directed representation learning. The survivor--skip structure of a constrained additive U-Net is shown to be exactly equivalent to a critically sampled multirate PR filter bank. The full-rate formulation removes the complementary-subband restrictions of the critically sampled system while preserving PR. A Residual Full-Rate PR architecture is then proposed to progressively route task-irrelevant, nuisance, or redundant structure away from the task-facing survivor while retaining the routed information explicitly. Exact reconstruction is guaranteed for arbitrary shape-compatible linear or nonlinear routing operators, without requiring invertibility, a matched synthesis bank, reconstruction loss, or learned decoder. The resulting architecture decouples representation design from reconstruction design: conservation is structural, while learning is devoted to task-directed routing. The same formulation identifies an identity-shortcut ResNet with its residual output retained as a full-rate PR system. Experiments verify exact single-channel routing of linearly separable factors to machine precision. On TIMIT, the proposed front-end improves test PER from $28.60\pm2.09\%$ to $25.76\pm0.41\%$ with the recognizer and training protocol held fixed, while maintaining exact reconstruction. Speaker probing further shows that structural conservation does not itself imply task-specific invariance.