Deep Learning in Quantitative Trading. 2025. Zihao Zhang and Stefan Zohren. Cambridge University Press.
The use of deep learning in finance has exploded over the last five years. However, the concepts and research have generally not received a comprehensive survey for portfolio managers and research analysts. Deep Learning in Quantitative Trading by Zihao Zhang and Stefan Zohren is not an investment finance textbook but a theoretical introduction with practical applications using deep learning tools. This work is the most recent offering in a series of short books called the Cambridge Elements in Quantitative Finance. The goal of the series is to present specific quantitative finance topics in a more detailed and accessible fashion than a research paper, while being shorter than a textbook. Although the reader requires quantitative skills, the series presents new and complex topics in a manner that is accessible to a finance audience wider than one for pure academic research.
This specialized book focuses on deep learning, a subset of machine learning that employs neural networks with multiple layers to learn features of and account for nonlinearity in raw data. Zhang and Zohren break their work into two parts: a foundation sector, which deals with the math and intuition behind deep learning, and an application sector, which focuses on how deep learning models can be used to solve many standard quant finance problems.
Instead of linear regression or decision trees, the authors concentrate on explaining different types of fully connected networks, including convolutional neural networks (CNNs), recurrent neural networks (RNNs) such as long short-term memory (LSTM) networks, and graphical neural networks (GNNs). The authors also discuss attention mechanisms and transformer networks, including encoder-decoder methodologies. All are useful for complex nonlinear prediction tasks, yet each employs a different approach to address the challenges posed by financial time series. While these neural networks are the workhorses employed by the authors, other forms of deep learning are reviewed to illustrate the richness of the choices available to an analyst.
The foundation portion of the book focuses on the fundamentals of financial time series, a presentation of supervised learning and canonical networks, and model training workflow. This sequential section explains how different neural networks can be helpful for specific problems so that the right deep learning model matches the appropriate problem. The approach of walking through the process of building models with a clear set of steps from data collection through exploratory data analysis, cross-validation techniques, and parameter tuning is much appreciated. A process-driven method significantly affects implementation and the results.
The application section presents comparative results across different modeling approaches for the same problem, highlighting the benefits of deep learning techniques. These practical comparisons are the key value of this work. A clear roadmap for model building and implementation, with data results across model types and timeframes, shows the marginal value of employing deep learning techniques. These application examples make a strong case for enhancing classical quantitative trading strategies with deep learning. The authors also show how these nonlinear techniques can improve risk management and portfolio optimization.
The application starts with the classic quant problem of building a time-series momentum model (trend-following) and a cross-sectional momentum model to form long–short portfolios. The alternative results displayed make it clear that even simple deep learning models can add value without overfitting. Of course, the results for a given strategy or optimization also point to the problem of dealing with complex and dynamic markets; nothing works all the time. Zhang and Zohren spend the last part of the applications section on market microstructure and high-frequency trading. The complexities of the market order book and trade flows make deep learning techniques especially useful, but solving these problems also requires deep domain knowledge.
The problem with any presentation of complex quantitative techniques is finding the right balance between rigor and review. The authors are generally successful in their presentation, with good references for deeper analysis and further reading. Yet, this work may still be difficult for readers without a strong quantitative foundation. At times, the book will touch on a topic only to jump to another after a quick introduction. The reader gets a taste of the technique but will have to look elsewhere for further depth. Nevertheless, the text and applications are closely integrated with GitHub programming examples and code scripts so the reader can seamlessly move between theory, explanations, and practical applications.
After analyzing the differences in techniques, it becomes clear that working closely with the theory behind deep learning techniques is worth the high barrier to entry. There is tangible value in investing in deep learning, although the ongoing challenge is explaining or interpreting results, an issue that needs closer analysis across all machine learning techniques.
The authors may underrepresent a key component of successful deep learning models, domain knowledge. While practical applications demonstrate how deep learning can solve specific problems, there is still a need to understand the asset space before applying these complex techniques. A greater focus on explaining deep learning techniques would have been preferred over the deep dive into market microstructure and high-frequency problems; nevertheless, the authors achieve the goal of making a complex topic accessible. The only negative is that figures and tables could be better described and self-contained to have a greater impact in furthering the key narratives.
The knowledge in this quantitative area of finance is rapidly evolving, and providing quick survey work to reach a broad audience is a critical demand by the finance profession. Practitioners need more well-written, comprehensive surveys specific to machine learning in finance. Two additional pieces that will support this goal are the working paper “Deep Learning in Asset Management: Architectures, Applications, and Challenges,” by Yoontae Hwang et al. (http://dx.doi.org/10.2139/ssrn.5593850) and the monograph published by CFA Institute Research Foundation and CFA Institute Research and Policy Center, AI in Asset Management: Tools, Applications, and Frontiers, edited by Joseph Simonian, PhD. The combination of Deep Learning in Quantitative Trading withthese other two surveys, along with selective use of the focused references cited in the book, will place any quantitative analyst or general portfolio manager in a position to effectively build or assess the use of deep learning in finance.
If you liked this post, don’t forget to subscribe to the Enterprising Investor.
All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.