Spiking Neural Network for Cross-Market Portfolio Optimization in Financial Markets: A Neuromorphic Computing Approach

TL;DR

This paper tackles cross-market portfolio optimization under realistic trading constraints by employing a Spiking Neural Network (SNN) framework on neuromorphic computing principles. It processes five years of daily data from the Indian Nifty 500 and US S&P 500, uses hierarchical clustering for dimensionality reduction, and employs population-based spike encoding, Leaky Integrate-and-Fire neurons with adaptive thresholds, spike-timing-dependent plasticity, and multi-objective optimization to maximize risk-adjusted returns under cardinality, transaction costs, and evolving risk aversion. Experimental results show that the SNN-based portfolio achieves higher risk-adjusted performance and lower volatility than a baseline ANN, while offering substantial computational advantages due to event-driven processing (e.g., ~78% reduction in active computations and real-time processing capability). The work underscores the potential of neuromorphic approaches for scalable, energy-efficient, and interpretable portfolio optimization across global markets and lays a foundation for further hardware-based deployment and extension to additional asset classes and markets.

Abstract

Cross-market portfolio optimization has become increasingly complex with the globalization of financial markets and the growth of high-frequency, multi-dimensional datasets. Traditional artificial neural networks, while effective in certain portfolio management tasks, often incur substantial computational overhead and lack the temporal processing capabilities required for large-scale, multi-market data. This study investigates the application of Spiking Neural Networks (SNNs) for cross-market portfolio optimization, leveraging neuromorphic computing principles to process equity data from both the Indian (Nifty 500) and US (S&P 500) markets. A five-year dataset comprising approximately 1,250 trading days of daily stock prices was systematically collected via the Yahoo Finance API. The proposed framework integrates Leaky Integrate-andFire neuron dynamics with adaptive thresholding, spike-timingdependent plasticity, and lateral inhibition to enable event-driven processing of financial time series. Dimensionality reduction is achieved through hierarchical clustering, while populationbased spike encoding and multiple decoding strategies support robust portfolio construction under realistic trading constraints, including cardinality limits, transaction costs, and adaptive risk aversion. Experimental evaluation demonstrates that the SNN-based framework delivers superior risk-adjusted returns and reduced volatility compared to ANN benchmarks, while substantially improving computational efficiency. These findings highlight the promise of neuromorphic computation for scalable, efficient, and robust portfolio optimization across global financial markets.

Figures (10)

• Figure 1: Traditional artificial neural network architecture showing fully connected layers with continuous activation functions. The network processes all input features simultaneously at each time step, resulting in high computational overhead for large-scale portfolio optimization problems.
• Figure 2: Proposed spiking neural network architecture incorporating leaky integrate-and-fire neurons, lateral inhibition connections, and spike-timing-dependent plasticity learning mechanisms. The event-driven processing enables computational efficiency through sparse activation patterns.
• Figure 3: Comprehensive data processing pipeline for cross-market portfolio optimization, showing the flow from raw financial data through preprocessing, spike encoding, neural network training, and final portfolio construction with constraint enforcement.
• Figure 4: Distribution of annualized mean returns and volatilities for the portfolio optimization dataset, showing the heterogeneous nature of risk-return characteristics across Indian and US equity markets with return distribution centered around 20-25% and volatility concentrated in the 20-40% range.
• Figure 5: Cumulative yearly return comparison between SNN and ANN optimized portfolios over the 2020-2025 evaluation period, demonstrating consistent outperformance of the neuromorphic approach with final returns of 42.74% versus 30.25% for traditional optimization.