公开(公告)号：US20250036940A1

公开(公告)日：2025-01-30

申请号：US18637161

申请日：2024-04-16

Applicant: Apple Inc.

Inventor： Ali MOIN ,  Ran LIU ,  Ellen L. ZIPPI ,  Mohammad Hadi POUR ANSARI ,  Christopher M. SANDINO ,  Erdrin AZEMI

IPC: G06N3/08

Abstract: The subject technology provides frequency-aware masked autoencoders for multimodal pretraining on frequency-based signals. An apparatus receives input data comprising frequency-based signal information associated with one or more modalities. The apparatus transforms the input data from a time domain to a frequency domain. The apparatus generates a frequency-embedded latent representation of the input data comprising time-domain and frequency-domain information. The apparatus also generates a masked frequency-embedded latent representation by masking one or more frequency components in the frequency-embedded latent representation. The apparatus produces a trained machine learning model by training a neural network to predict one or more masked frequency components of the frequency-embedded latent representation.