other

multivariate time series representation learning

otheractiveprovisionalmultivariate-time-series-representation-learning-c74250aa·1 events·first seen 16d ago

Aliases: multivariate time series representation learning

Co-occurring entities

Transformers CHARM Joint Embedding Predictive Architecture (JEPA)description-aware gating

More like this (12)

time-series forecasting Time Series Foundation Models Disentangled Representation Learning Beyond Uniform Tokens: Adaptive Compression for Time Series Language Models multimodal classification models Multimodal Learning A Diffusion Approximation for Temporal-Difference Learning with Linear Features under Markovian Noise When to Align, When to Predict: A Phase Diagram for Multimodal Learning Matryoshka Representation Learning Probabilistic Time Series Forecasting Latent World Recovery for Multimodal Learning with Missing Modalities Multi-Task Learning

Recent events (1)

6arXiv · cs.LG·16d ago·source ↗

CHARM: Multimodal JEPA for Semantic Time-Series Embeddings via Channel-Aware Representation Learning

CHARM (Channel-Aware Representation Model) is a new Transformer-based architecture for general-purpose representation learning over heterogeneous multivariate time series. It integrates channel-level textual descriptions into a permutation-equivariant encoder trained with a Joint Embedding Predictive Architecture (JEPA) and a novel temporally stable embedding loss. The model achieves strong performance across anomaly detection, classification, and forecasting tasks using only a linear probe, with text descriptions primarily serving as channel identifiers enabling cross-dataset generalization.

Agent and Tool Ecosystem Multimodal Progress Transformers CHARM Joint Embedding Predictive Architecture (JEPA)+2 more