Abstract:We present a neural network based calibration method that performs the calibration task within a few milliseconds for the full implied volatility surface. The framework is consistently applicable throughout a range of volatility models -including the rough volatility family- and a range of derivative contracts. The aim of neural networks in this work is an off-line approximation of complex pricing functions, which are difficult to represent or time-consuming to evaluate by other means. We highlight how this perspective opens new horizons for quantitative modelling: The calibration bottleneck posed by a slow pricing of derivative contracts is lifted. This brings several numerical pricers and model families (such as rough volatility models) within the scope of applicability in industry practice. The form in which information from available data is extracted and stored influences network performance: This approach is inspired by representing the implied volatility and option prices as a collection of pixels. In a number of applications we demonstrate the prowess of this modelling approach regarding accuracy, speed, robustness and generality and also its potentials towards model recognition.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is: **How to accelerate the pricing and calibration processes of volatility models (especially rough volatility models) through deep - learning methods, thereby overcoming the bottleneck of slow speed in traditional numerical methods**.
Specifically, the paper proposes a calibration method based on neural networks, which can complete the calibration of the complete implied volatility surface within a few milliseconds. This method is applicable to a series of volatility models (including rough volatility models) and various derivative contracts. The following is a summary of the core content of the paper:
### 1. **Problem Background**
- **Challenges of Volatility Models**: Many volatility models (such as rough volatility models) have long computational times in traditional pricing methods (such as Monte Carlo simulation) due to their complexity, resulting in a very time - consuming calibration process.
- **Limitations in Industrial Applications**: This computational bottleneck limits the practical applications of these models, despite their many theoretical advantages.
### 2. **Solution**
- **Application of Neural Networks**: The paper proposes using deep neural networks to approximate complex pricing functions. Neural networks can be trained in the offline stage to learn the mapping relationship from model parameters to option prices.
- **Fast Calibration**: Once trained, the neural network can output the implied volatility surface corresponding to the given model parameters within a few milliseconds, greatly improving the calibration speed.
- **Generality and Accuracy**: This method is not only fast but also shows good generalization ability and high accuracy on different models and parameter combinations.
### 3. **Technical Details**
- **Neural Network Architecture**: The paper describes in detail the structural design of the neural network, including the specific settings of the input layer, hidden layer, and output layer. The input is model parameters, and the output is the corresponding implied volatility or option price.
- **Training Process**: Use sample data generated by traditional numerical methods as a training set, and train the neural network through supervised learning so that it can accurately approximate the pricing function.
- **Calibration Steps**: In the calibration stage, use the trained neural network to quickly calculate the implied volatility surface and compare it with market data, and adjust the model parameters to minimize the error.
### 4. **Experimental Results**
- **Numerical Experiments**: The paper shows the results of multiple numerical experiments, verifying the accuracy and speed of this method under different models (such as the rough Bergomi model).
- **Historical Data Calibration**: It also conducts calibration experiments based on historical data, showing the performance of this method on actual market data.
### 5. **Conclusion and Outlook**
- **Potential Applications**: This method is not only applicable to the calibration of volatility models but can also be extended to other financial engineering fields, such as the pricing of complex derivatives and risk management.
- **Future Work**: Further explore the application of neural networks in more types of models and optimize the network structure and training algorithms.
### Formula Presentation
Some formulas involved in the paper are presented in Markdown format as follows:
- **Pricing Mapping**:
\[
P: M(\theta, \zeta) \to \mathbb{R}^m
\]
where \( \theta \in \Theta \) is a model parameter, \( \zeta \) is a financial product (such as a European option), and \( m \) is the number of options.
- **Calibration Problem**:
\[
\hat{\theta} = \arg\min_{\theta \in \Theta} \delta(P(M(\theta), \zeta), P_{\text{MKT}}(\zeta))
\]
where \( \delta(\cdot, \cdot) \) is a suitable metric function and \( P_{\text{MKT}}(\zeta) \) is the market - observed price.
- **Neural Network Approximation**:
\[
\tilde{F}: \Theta \to \tilde{P}(M)
\]
\[
\tilde{F}(\theta) = P(M(\theta), \zeta) + O(\epsilon)
\]
Through these improvements, the paper has successfully solved.