deepecho

What it is and what it does

DeepEcho is a Python library for generating synthetic time series data with mixed data types (categorical, continuous, count). It combines classical statistical modeling with deep learning techniques—specifically GANs—to learn patterns from your data and produce new sequences that preserve the statistical properties of the original. The library is part of the Synthetic Data Vault (SDV) project but can be used standalone.

You provide your time series data, define column types and structural metadata (entity columns, context columns, sequence index), and the model learns to generate new synthetic sequences. It supports both CPU and GPU training, though the overhead of GPU may not be worth it on small datasets. The library is actively maintained but still in Pre-Alpha, so expect potential API changes.

Use it for:

• Generate synthetic time series for testing machine learning pipelines without exposing real customer or operational data.
• Augment small time series datasets to improve training data volume for downstream models.
• Create benchmark datasets for evaluating time series forecasting or anomaly detection algorithms.
• Simulate realistic multivariate sequences for privacy-preserving data sharing in regulated industries.
• Prototype and validate time series analysis workflows before deploying on production data.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

DeepEcho generates synthetic time series data with mixed data types using deep learning and statistical models, supporting both standalone use and integration with the broader SDV ecosystem.

Yes, with conditions. Install if you need to generate synthetic time series with mixed data types and can accept Pre-Alpha stability. The low install friction, active maintenance, and zero known vulnerabilities are positive signals. However, review the BUSL-1.1 license terms carefully—commercial use is restricted until a specified date. If you are in research, internal testing, or non-commercial use, this is a reasonable choice; if commercial deployment is your goal, verify the license terms or contact DataCebo.

Install

deepecho on PyPI

pip install deepecho

uv add deepecho

poetry add deepecho

Installing deepecho

Before you install

Low install friction with a pure-Python wheel. Active maintenance with recent commits. Depends on numpy, pandas, torch, and tqdm—all standard, well-maintained libraries. Pre-Alpha status means the API may change.

License in practice

Licensed under BUSL-1.1 (Business Source License), which restricts commercial use until a future date and requires review of terms before deployment in production or commercial contexts.

Quickstart

pip install deepecho

from deepecho import PARModel
from deepecho.demo import load_demo

data = load_demo()
model = PARModel(cuda=False)
model.fit(data=data, entity_columns=['store_id'], context_columns=['region'], data_types={'region': 'categorical', 'total_sales': 'continuous'}, sequence_index='date')
model.sample(num_entities=5)

Requires PyTorch (torch) as a runtime dependency; CUDA is optional but may add overhead on small datasets.

Verify before relying

• Whether BUSL-1.1 restrictions apply to your intended use case (commercial, research, or internal).
• Specific performance or quality benchmarks compared to other time series synthesis methods.
• Whether the Pre-Alpha status means breaking changes are expected in the near term.

Package facts

Evidence: deepecho-0.8.1-py3-none-any.whl

Keywords: deepecho, DeepEcho

Tags

More Artificial Intelligence packages