Rotary Masked Autoencoders are Versatile Learners

Uros Zivanovic, Serafina Di Gioia, Andre Scaffidi, Martín Emilio de los Rios, Gabriella Contardo, Roberto Trotta

Keywords: deep learning, Transformer, Masked Autoencoder, irregular time-series, Rotary Positional Embeddings

TL;DR: RoMAE is a single, drop-in masked autoencoder that utilizes continuous (axial) RoPE to excel in interpolation and representation learning across modalities.

Abstract: Applying Transformers to irregular time-series typically requires specializations to their baseline architecture, which can result in additional computational overhead and increased method complexity. We present the Rotary Masked Autoencoder (RoMAE), which utilizes the popular Rotary Positional Embedding (RoPE) method for continuous positions. RoMAE is an extension to the Masked Autoencoder (MAE) that enables interpolation and representation learning with multidimensional continuous positional information while avoiding any time-series-specific architectural specializations. We showcase RoMAE's performance on a variety of modalities including irregular and multivariate time-series, images, and audio, demonstrating that RoMAE surpasses specialized time-series architectures on difficult datasets such as the DESC ELAsTiCC Challenge while maintaining MAE's usual performance across other modalities. In addition, we investigate RoMAE's ability to reconstruct the embedded continuous positions, demonstrating that including learned embeddings in the input sequence breaks RoPE's relative position property.

Primary Area: Deep learning (e.g., architectures, generative models, optimization for deep networks, foundation models, LLMs)

Submission Number: 23740