Course Program

Purpose of Lectures 

The DEEP Learning Summer School 2026 is designed to equip PhD students in economics, finance, and related fields with the cutting-edge computational tools that are rapidly transforming modern economic and financial research. By integrating applied mathematics, machine learning, and computational economics, these lectures provide a deep dive into two highly complementary frontiers: solving and estimating complex dynamic stochastic models, and leveraging Artificial Intelligence (AI), Natural Language Processing (NLP), and Large Language Models (LLMs) to tackle empirical questions in finance.

Core Pillars of the Program

Deep Learning for Solving and Estimating Dynamic Stochastic Models (Days 1 & 2)

The first half of the course focuses on using advanced neural network architectures to solve high-dimensional economic problems. Through theoretical foundations and hands-on applications, participants will explore:

• • Deep Equilibrium Nets & Surrogates: Introduction to solving dynamic stochastic models, complete with guided coding exercises, followed by utilizing Deep Surrogates for estimation.

  • Heterogeneous Agent Models: Demonstrating how machine learning, specifically DeepHAM and structural reinforcement learning, can efficiently solve complex models that were previously computationally intractable.

  • PDEs & Continuous-Time Models: Leveraging Physics-Informed Neural Nets (PINNs) to solve Partial Differential Equations, alongside dedicated deep learning techniques for continuous-time frameworks.

Large Language Models and Sequence Modeling for Economics and Finance (Days 2 & 3)

The second half of the course shifts to the processing of unstructured data, taking students on a comprehensive journey from foundational mathematics to modern autonomous agents:

• • Foundations of Sequence Modelling: Exploring the mathematical backbone of time series and language, from RNNs and LSTMs to the revolutionary Transformer architecture.

  • The LLM Revolution: Examining the pre-training paradigm, BERT and GPT architectures, scaling laws, and the emergent abilities of massive models.

  • Domain Adaptation: Addressing how to adapt general-purpose models for specific financial and climate intelligence using parameter-efficient fine-tuning (LoRA) and human-intent alignment (RLHF/DPO).

  • Advanced Frontiers & Applications: Moving beyond basic prompting to Retrieval-Augmented Generation (RAG), Agentic LLMs (planning and tool use), and a culminating look at practical applications of deep sequence modelling in finance.

Workshop Format & Hands-On Learning

What sets this summer school apart is its interactive, workshop-style format, complemented by insights from an upcoming Industry Talk. Participants will not only engage in rigorous theoretical discussions but will also apply their knowledge immediately through guided, hands-on coding exercises. With practical examples provided in Python, students will gain firsthand experience in implementing machine learning solutions, ranging from solving continuous-time models to extracting financial sentiment from text. All hands-on components will be powered by Nuvolos, a cloud-based teaching platform enabling reproducible, scalable, and frictionless computational environments.

Expected Outcomes

By the end of the series, participants will have developed a robust, dual skill set. Whether you are exploring macroeconomic policy simulations, estimating heterogeneous agent models, or utilizing modern LLMs to extract actionable intelligence from corporate disclosures, this summer school provides the essential, state-of-the-art tools to drive innovation in your research.

Registration Deadline: 30th April 2026

The schedule (To be confirmed)