Datasets:
Formats:
json
Languages:
English
Size:
< 1K
Tags:
time-series
time-series-decomposition
benchmark
component-recovery
symbolic-regression
icml-2026
License:
File size: 3,063 Bytes
17b7ba4 910431f 17b7ba4 910431f 17b7ba4 910431f 17b7ba4 910431f 17b7ba4 910431f 3b377aa 910431f 3b377aa 910431f 3b377aa 910431f 3b377aa 17b7ba4 910431f 3b377aa 910431f 17b7ba4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 | # TSDecompose Source Snapshot
This folder contains the source snapshot shipped with the benchmark release.
It is intentionally included as repository source code rather than as a PyPI-style package release. The import name remains `tsdecomp` for compatibility with the benchmark scripts.
## Layout
```text
src/tsdecomp/
Unified decomposition API, method registry, CLI, metrics, and leaderboard runner.
src/synthetic_ts_bench/
Synthetic scenario generator and component-recovery utilities.
scripts/run_paper_benchmark.py
One-command runner for the paper-aligned core benchmark.
requirements.txt
Runtime dependencies for local execution.
```
## One-Command Paper Benchmark
Install dependencies from this folder, then run the paper core benchmark:
```bash
pip install -r requirements.txt
export PYTHONPATH="$PWD/src"
python scripts/run_paper_benchmark.py
```
The script pins imports to this source snapshot, so it is safe to use even if a
different `tsdecomp` package is installed elsewhere on the same machine.
Windows PowerShell:
```powershell
pip install -r requirements.txt
$env:PYTHONPATH = "$PWD\src"
python scripts/run_paper_benchmark.py
```
The default command runs the paper-aligned core synthetic benchmark:
- 6 scenarios;
- 50 deterministic generated draws per scenario;
- 300 generated synthetic series total;
- the camera-ready six-family Table 2 roster (`ma_baseline`, `stl`, `ssa`, `emd`, `vmd`, `wavelet`).
The raw result table is written to:
```text
artifacts/paper_core_benchmark/leaderboard.csv
```
In the dataset release, this same paper-core raw output is stored as
`data/paper_tables/paper_core_50draw_leaderboard.csv` after stable row sorting.
The manuscript Table 2 values are derived from
`data/paper_tables/paper_core_50draw_by_tier.csv` using the tier-balanced rule:
stationary columns average Tier 1 and Tier 2 means with equal weight, while
non-stationary columns use Tier 3 means. The by-tier file records
metric-specific valid-row counts because seasonal metrics are undefined for the
trend-only scenario.
Aggregated summaries are written under:
```text
artifacts/paper_core_benchmark/summary/
```
For a quick dependency and integration check:
```bash
python scripts/run_paper_benchmark.py --smoke
```
This smoke run keeps the same 6 paper scenarios but uses 1 draw per scenario
and the smaller `stl,wavelet` method subset.
## Optional Direct CLI
```bash
python -m tsdecomp validate --suite core --methods ma_baseline,stl,ssa,emd,vmd,wavelet
```
Windows PowerShell:
```powershell
python -m tsdecomp validate --suite core --methods ma_baseline,stl,ssa,emd,vmd,wavelet
```
Use the script entrypoint above for reproduction. The direct module CLI is
shown only for clean environments where no other editable `tsdecomp` install can
shadow this source tree.
The native extension binary from the development machine is not included. The release uses the pure-Python fallback path by default. To test a separately built native extension, set `TSDECOMP_ALLOW_EXTERNAL_NATIVE=1` before importing `tsdecomp`.
|