Interpretable Convolutional Filters with SincNet

By Mirco Ravanelli, Yoshua Bengio. 2018

119 citations recorded by the source corpus.

Deep learning is currently playing a crucial role toward higher levels of artificial intelligence. This paradigm allows neural networks to learn complex and abstract representations, that are progressively obtained by combining simpler ones. Nevertheless, the internal "black-box" representations automatically discovered by current neural architectures often suffer from a lack of interpretability, making of primary interest the study of explainable machine learning techniques. This paper summarizes our recent efforts to develop a more interpretable neural model for directly processing speech from the raw waveform. In particular, we propose SincNet, a novel Convolutional Neural Network (CNN) that encourages the first layer to discover more meaningful filters by exploiting parametrized sinc functions. In contrast to standard CNNs, which learn all the elements of each filter, only low and high cutoff frequencies of band-pass filters are directly learned from data. This inductive bias offers a very compact way to derive a customized filter-bank front-end, that only depends on some parameters with a clear physical meaning. Our experiments, conducted on both speaker and speech recognition, show that the proposed architecture converges faster, performs better, and is more interpretable than standard CNNs.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety, Robustness & Interpretability

In Signal Processing

Citations in the corpus

Multi-Task Self-Supervised Learning for Robust Speech Recognition (2020, positive): “Similarly to [15], the first layer of the encoder is based on SincNet [22], which performs the convolution of the raw input waveform with a set of parameterized sinc functions implementing rectangular band-pass filters.”
Learning Problem-agnostic Speech Representations from Multiple Self-supervised Tasks (2019, methodology): “PASE-FineTuned also outperforms our previous results obtained with the standard SincNet model [22].”
Speech Model Pre-training for End-to-End Spoken Language Understanding (2019, methodology): “The phoneme module is implemented using a SincNet layer [41, 42], which processes the raw input waveform, followed by multiple convolutional layers and recurrent layers with pooling (to reduce the sequence length) and dropout.”
The Pytorch-kaldi Speech Recognition Toolkit (2018, methodology): “The SincNet model [33, 34] is also implemented to perform speech recognition from raw waveform directly.”
BENDR: Using Transformers and a Contrastive Self-Supervised Learning Task to Learn From Massive Amounts of EEG Data (2021, background): “This may provide insight into better weight initialization, or integration of explicit early layers similar to [6] (one could also argue that SincNet layers [71] are some such layers that could factor here).”
Deep Learning Algorithms for Rotating Machinery Intelligent Diagnosis: An Open Source Benchmark Study (2020, methodology): “of DL-based models from the following aspects: (1) visualize the results of neurons to analyze the attention points of models [62]; (2) add physical constraints to the loss function [63] to meet specific needs of fault feature extraction; (3) add prior knowledge to network structures and convolution”
Learning Speaker Representations with Mutual Information (2018, methodology): “To better process raw audio, the encoder is based on SincNet [21, 22], a novel Convolutional Neural Network (CNN) that encourages the first layer to discover more meaningful filters.”