“Paper 3 at the June 2023 ECB Conference: Exploring Forecasting Techniques”

**ECB Conference on Forecasting Techniques, June 2023 – Enhanced Bayesian Neural Networks for Macroeconomics and Finance**

Welcome to the ECB Conference on Forecasting Techniques! This biennial conference brings experts together to share new ideas on economic forecasting. The upcoming edition will focus on current challenges faced by forecasters, such as modeling economic dynamics after extreme events, risk assessment, and inflation forecasting.

***Conference Details***
Date: June 2023
Location: Online (Central European Time)
Conference Program: [Click here](https://www.ecb.europa.eu/pub/conferences/html/20230612_12th_conference_on_forecasting_techniques.en.html) for the detailed program.

This video discusses the use of Enhanced Bayesian Neural Networks for macroeconomics and finance. The presenters are experts from the University of Salzburg and they will explore the difficulties in forecasting macroeconomic and financial variables, especially in times of high uncertainty and risk. They will also discuss how neural networks can be used to overcome challenges such as non-linearities and modeling complexities.

The video highlights their innovative approach of using recent advances in Bayesian statistics and econometrics to determine the structure of the neural network, reducing the need for time-consuming cross-validation exercises. They showcase the effectiveness of their approach through simulations and the application of their model to macroeconomic and finance datasets.

**Key Topics Covered in this Video:**
– Challenges in forecasting macroeconomic and financial variables
– Introduction to neural networks for learning unknown relationships between variables
– The drawback of model specification in neural networks
– Innovative use of Bayesian statistics and econometrics to determine network structure
– Advantages of using Bayesian neural networks for safeguarding against overfitting
– Model selection using shrinkage priors and multiplicative gamma process

**Transcript Excerpt:**
“In our research, we focus on enhancing the capabilities of Bayesian neural networks for macroeconomics and finance. Forecasting variables such as inflation, industrial reaction, and stock market indices has always been a challenging task, especially during uncertain times like the recent pandemic.

To address these challenges, we propose the use of neural networks as a device for learning relationships between variables. However, the task of model specification in neural networks can be quite complex, with a large number of hyperparameters to choose from. In our approach, we leverage recent advancements in Bayesian statistics and econometrics to determine the structure of the network, reducing the need for time-consuming cross-validation exercises.

Our method not only addresses issues of overfitting but also allows for the introduction of stochastic volatilities in the model. Empirically, we have demonstrated the effectiveness of our approach through simulations and its application to prominent macroeconomic and finance datasets.

This video provides a detailed explanation of our model, focusing on the use of shallow neural networks. We also extend our approach to deep neural networks in our research. Our contribution lies in the innovative use of shrinkage priors and a multiplicative gamma process for model selection.

Join us in this informative video as we delve into the enhanced Bayesian neural networks for macroeconomics and finance!”

**Sources:**
– [ECB Conference on Forecasting Techniques](https://www.ecb.europa.eu/pub/conferences/html/20230612_12th_conference_on_forecasting_techniques.en.html)
– [University of Salzburg](https://www.uni-salzburg.at/)
– [Bayesian Neural Networks](https://en.wikipedia.org/wiki/Bayesian_neural_network)

ECB Conference on Forecasting Techniques, June 2023 – Enhanced Bayesian Neural Networks for Macroeconomics and Finance.

The biennial ECB Conference of Forecasting Techniques provides a forum for new theoretical and applied work on economic forecasting. The forthcoming edition will bring together experts to exchange new ideas on some of the main current challenges faced by forecasters including the modelling of economic dynamics after extreme events, the assessment of risks, and inflation forecasting.

Programme Times are Central European Time * indicates the presenter
https://www.ecb.europa.eu/pub/conferences/html/20230612_12th_conference_on_forecasting_techniques.en.html