On the Constrained Time-Series Generation Problem
Authors: Andrea Coletta, Sriram Gopalakrishnan, Daniel Borrajo, Svitlana Vyetrenko
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically evaluate our work on several datasets for financial and energy data, where incorporating constraints is critical. We show that our approaches outperform existing work both qualitatively and quantitatively, and that Guided Diff Time does not require re-training for new constraints, resulting in a significant carbon footprint reduction, up to 92% w.r.t. existing deep learning methods. |
| Researcher Affiliation | Industry | Andrea Coletta J.P. Morgan AI Research London, UK Sriram Gopalakrishan J.P. Morgan AI Research New York, USA Daniel Borrajo J.P. Morgan AI Research Madrid, ESP Svitlana Vyetrenko J.P. Morgan AI Research New York, USA |
| Pseudocode | Yes | Algorithm 1 Guided-Diff Time Input: differentiable constraint fc : X R, scale parameter ρ Output: new TS, x0 x T sample from N(0, I) for all t from T to 1 do ˆϵ ϵθ(xt, t) ˆϵ ˆϵ ρ 1 ˆαt xtfc( 1 ˆαt (xt ˆϵ 1 ˆαt)) ˆαt 1 xt 1 ˆαtˆϵ ˆαt 1 ˆαt 1ˆϵ end for return x0 |
| Open Source Code | No | The paper does not contain an explicit statement about the release of its source code or a direct link to a code repository. |
| Open Datasets | Yes | Datasets We consider three datasets with different characteristics such as periodicity, noise, correlation, and number of features: 1) daily stocks which uses daily historical Google stock data from 2004 to 2019 with open, high, low, close, adjusted close, and volume features [1]; 2) energy data from the UCI Appliances energy prediction dataset [46] containing 28 features with noisy periodicity and correlation; 3) sines a synthetic multivariate sinusoidal TS with different frequencies and phases [1]. |
| Dataset Splits | No | The paper does not explicitly provide details about the training, validation, and test splits for the datasets used to train the generative models themselves. It describes how evaluation metrics are computed using post-hoc models trained on synthetic data. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | Yes | In our experiments, we use the Sequential Least Squares Programming (SLSQP) solver [27] in Scipy s optimization module [33]. [33] Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J. van der Walt, Matthew Brett, Joshua Wilson, K. Jarrod Millman, Nikolay Mayorov, Andrew R. J. Nelson, Eric Jones, Robert Kern, Eric Larson, C J Carey, Ilhan Polat, Yu Feng, Eric W. Moore, Jake Vander Plas, Denis Laxalde, Josef Perktold, Robert Cimrman, Ian Henriksen, E. A. Quintero, Charles R. Harris, Anne M. Archibald, Antônio H. Ribeiro, Fabian Pedregosa, Paul van Mulbregt, and Sci Py 1.0 Contributors. Sci Py 1.0: Fundamental Algorithms for Scientific Computing in Python. Nature Methods, 17:261 272, 2020. |
| Experiment Setup | Yes | If the COP solver cannot find a solution within the allowed error tolerance, we double the budget and repeat the process for up to a fixed number of η repeats (we set η = 10 in our experiments). Where b is the budget we set (b = 0.1 in our experiments). We set σt = 0, t [0, T] to have a deterministic forward process from latent variables to the sample x0 (since the noise term σtϵ is zeroed out). |