Speaker
Description
In this RF Machine Learning talk, we introduce the Sig53 dataset, the open-source TorchSig toolkit, and competitive baselines for the task of signal classification using the newly introduced dataset. The Sig53 dataset is a narrowband signals dataset consisting of 5 million synthetically generated samples from 53 different signal classes and expertly chosen impairments meant to accelerate research in the task of signal classification. TorchSig is an open-source signal processing machine learning toolkit with the ability to generate the Sig53 dataset, apply over 50 domain-tailored data augmentations/transformations, interface with neural networks through an image-domain inspired model API, and provide numerous data processing utilities meant to accelerate research in the broad space of RF machine learning. We also share competitive baseline performances of convolutional neural networks and transformer-based architectures on the task of narrowband signal classification using the Sig53 dastaset.
| Talk Length | 30 Minutes |
|---|---|
| Acknowledge | Acknowledge In-Person |
