Streaming Anchor Loss: Augmenting Supervision with Temporal Significance

TL;DR

This paper introduces Streaming Anchor Loss (SAL), a novel loss function that emphasizes learning from critical frames in streaming models, improving accuracy and latency without extra data or parameters.

Contribution

The paper proposes SAL, a dynamic loss that enhances learning focus on important frames in streaming models, addressing resource constraints.

Findings

SAL improves task accuracy in streaming speech models.

SAL reduces latency without increasing model complexity.

Experimental results confirm SAL's effectiveness across multiple tasks.

Abstract

Streaming neural network models for fast frame-wise responses to various speech and sensory signals are widely adopted on resource-constrained platforms. Hence, increasing the learning capacity of such streaming models (i.e., by adding more parameters) to improve the predictive power may not be viable for real-world tasks. In this work, we propose a new loss, Streaming Anchor Loss (SAL), to better utilize the given learning capacity by encouraging the model to learn more from essential frames. More specifically, our SAL and its focal variations dynamically modulate the frame-wise cross entropy loss based on the importance of the corresponding frames so that a higher loss penalty is assigned for frames within the temporal proximity of semantically critical events. Therefore, our loss ensures that the model training focuses on predicting the relatively rare but task-relevant frames. Experimental results with standard lightweight convolutional and recurrent streaming networks on three different speech based detection tasks demonstrate that SAL enables the model to learn the overall task more effectively with improved accuracy and latency, without any additional data, model parameters, or architectural changes.