Signature Methods for Time Series Data
summary
Signatures turn irregular, oscillatory paths into fixed‑size feature vectors and the pipeline is concrete: compute signatures with e‑cig stream_to_sig and signatory (PyTorch), truncate at a chosen level (example: depth‑8 signature = 511 features), average signatures to get class prototypes, and compare with a distance (example: max‑modulus distance 5.27 separated a three from a six). Implementations run on MNIST strokes, MIMIC three sepsis detection, action/drone/NLP tasks and show practical dimensionality compression and useful classification performance.
AI-written summary · about
chapters (12)
- 0:00Start
- 2:28Applying Math in Python
- 5:32Control Differential Equations
- 7:21The Signature of the Path
- 7:54What Is a Signature
- 13:05Invariant to Reparameterization
- 29:58Python Packages for Computing Signatures
- 33:17Handwriting Recognition
- 33:43Sepsis Detection
- 34:07Human Action Recognition
- 36:09Where To Find Out More
- 40:37Fourier Transform
related talks
19mUsing Modes for Time Series Classification
35mSignatures, patterns and trends: Timeseries data mining @ Etsy
28mSignature Verification in Banks using Few Shot Learning
27mTaming the Wild West of ML: Practical Model Signing with Sigstore on Kaggle
26mTaming ML’s Wild West: Practical Model Signing with Sigstore on Kaggle
21mML Model Signing: Cryptographically Paving the Way to Provenance in Machine Learning Models
55mSearching for Time Series Patterns at Big Data Scale
19mWe Know Security but How Do We Secure GenAI End-to-End?