Research map / Time series / Causal and generative time series

Kernel Representation Learning for Time Sequence: Algorithm, Theory, and Applications

AAAI 2025 · Time series › Causal and generative time series · approach: Generative temporal point processes · 3 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

A doctoral consortium thesis proposal outlining Kernel Representation Learning (KRL) for time sequences, summarizing completed work on adaptive kernel learning, kernel self-representation for co-evolving series, and planned theory/application/synthetic-data directions.

Problem

Modeling nonlinear, noisy, co-evolving multivariate time series with concept drift while balancing predictive accuracy and interpretability.

Approach

Use kernel-based self-representation learning to capture temporal dependencies and hidden structure in co-evolving series as an ecosystem, detect concept drift, and connect KRL theoretically to Transformers and Kolmogorov-Arnold Networks.

Key results

• Summarizes prior works: data-driven kernel learning and multiple kernel learning for adaptive kernel representations (SRQ.1), a kernel induced self-representation model for co-evolving series and concept drift (SRQ.2), and partial result that the Transformer kernel can be approximated as a dot produ

Limitations

• Mostly a research plan; several sub-questions (SRQ.3, SRQ.5, SRQ.6) are partial or not completed.
• No quantitative experimental results reported in the extract.
• Synthetic dataset generation and survey are future work.

Experimental setup

Builds on: multiple kernel learning self-attention Kolmogorov-Arnold Networks

Closest papers by meaning

• Causal Temporal Representation Learning with Nonstationary Sparse Transition (NeurIPS 2024 · similarity 0.86)
• Segment, Shuffle, and Stitch: A Simple Layer for Improving Time-Series Representations (NeurIPS 2024 · similarity 0.86)
• T-Rep: Representation Learning for Time Series using Time-Embeddings (ICLR 2024 · similarity 0.85)
• STEP: Learning STructured Embeddings for Progressive Time Series (NeurIPS 2026 · similarity 0.85)
• Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning (AAAI 2025 · similarity 0.85)
• Language Pre-training Guided Masking Representation Learning for Time Series Classification (AAAI 2025 · similarity 0.85)