Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Modal Financial Time-Series Retrieval Through Latent Space Projections

Tom Bamford, Andrea Coletta, Elizabeth Fons, Sriram Gopalakrishnan, Svitlana Vyetrenko, Tucker Balch, Manuela Veloso

TL;DR

The paper tackles the challenge of efficiently retrieving financial time-series (TS) when users want more flexible queries than fixed SQL-like expressions. It introduces a deep-encoder framework that learns a shared latent space across modalities (text, sketches, and TS-derived images) to store and retrieve TS while preserving finance-relevant properties such as volatility. The approach combines a CLIP-like text–image alignment with a sketch-enabled autoencoder pipeline, indexed via FAISS for fast lookup, and is evaluated on synthetic and real historical data, showing improved retrieval speed and competitive accuracy against baselines. The work demonstrates the practicality of latent-space projections for multi-modal TS storage and retrieval and highlights its potential to enable intuitive, scalable querying in financial analytics.

Abstract

Financial firms commonly process and store billions of time-series data, generated continuously and at a high frequency. To support efficient data storage and retrieval, specialized time-series databases and systems have emerged. These databases support indexing and querying of time-series by a constrained Structured Query Language(SQL)-like format to enable queries like "Stocks with monthly price returns greater than 5%", and expressed in rigid formats. However, such queries do not capture the intrinsic complexity of high dimensional time-series data, which can often be better described by images or language (e.g., "A stock in low volatility regime"). Moreover, the required storage, computational time, and retrieval complexity to search in the time-series space are often non-trivial. In this paper, we propose and demonstrate a framework to store multi-modal data for financial time-series in a lower-dimensional latent space using deep encoders, such that the latent space projections capture not only the time series trends but also other desirable information or properties of the financial time-series data (such as price volatility). Moreover, our approach allows user-friendly query interfaces, enabling natural language text or sketches of time-series, for which we have developed intuitive interfaces. We demonstrate the advantages of our method in terms of computational efficiency and accuracy on real historical data as well as synthetic data, and highlight the utility of latent-space projections in the storage and retrieval of financial time-series data with intuitive query modalities.

Multi-Modal Financial Time-Series Retrieval Through Latent Space Projections

TL;DR

The paper tackles the challenge of efficiently retrieving financial time-series (TS) when users want more flexible queries than fixed SQL-like expressions. It introduces a deep-encoder framework that learns a shared latent space across modalities (text, sketches, and TS-derived images) to store and retrieve TS while preserving finance-relevant properties such as volatility. The approach combines a CLIP-like text–image alignment with a sketch-enabled autoencoder pipeline, indexed via FAISS for fast lookup, and is evaluated on synthetic and real historical data, showing improved retrieval speed and competitive accuracy against baselines. The work demonstrates the practicality of latent-space projections for multi-modal TS storage and retrieval and highlights its potential to enable intuitive, scalable querying in financial analytics.

Abstract

Financial firms commonly process and store billions of time-series data, generated continuously and at a high frequency. To support efficient data storage and retrieval, specialized time-series databases and systems have emerged. These databases support indexing and querying of time-series by a constrained Structured Query Language(SQL)-like format to enable queries like "Stocks with monthly price returns greater than 5%", and expressed in rigid formats. However, such queries do not capture the intrinsic complexity of high dimensional time-series data, which can often be better described by images or language (e.g., "A stock in low volatility regime"). Moreover, the required storage, computational time, and retrieval complexity to search in the time-series space are often non-trivial. In this paper, we propose and demonstrate a framework to store multi-modal data for financial time-series in a lower-dimensional latent space using deep encoders, such that the latent space projections capture not only the time series trends but also other desirable information or properties of the financial time-series data (such as price volatility). Moreover, our approach allows user-friendly query interfaces, enabling natural language text or sketches of time-series, for which we have developed intuitive interfaces. We demonstrate the advantages of our method in terms of computational efficiency and accuracy on real historical data as well as synthetic data, and highlight the utility of latent-space projections in the storage and retrieval of financial time-series data with intuitive query modalities.
Paper Structure (34 sections, 7 equations, 2 figures, 3 tables)

This paper contains 34 sections, 7 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Commercial and open-source TS databases.
  4. Research Literature
  5. Problem Description
  6. Dataset Construction
  7. Synthetic stock price time series dataset
  8. Unfiltered
  9. Filtered
  10. Filtered+
  11. Historical price time series dataset
  12. Methodologies
  13. Text-based retrieval
  14. Architecture
  15. Image Encoder
  16. ...and 19 more sections

Figures (2)

  • Figure 1: Interfaces developed for natural language and sketch based queries.
  • Figure 2: General model architecture for both retrieval methods. The Deep Encoders are trained with different modes of data against the loss to create a shared latent-space. A database containing the TS and the latent-space indexing is created. Finally, the user query is converted through the trained encoder into the latent space to query the database, and matching TS are retrieved.