Syllabus

Course Information

Course Title: Neural Networks - From Brain to Business

Description: This course introduces neural networks from biological inspiration through to practical business applications in financial prediction. Students will understand how artificial neurons work, how networks learn, and how to evaluate their performance.

Network Architecture

Prerequisites:

  • Basic algebra (variables, equations, functions)
  • Basic statistics (mean, probability, standard deviation)
  • No programming experience required

Learning Objectives

Upon successful completion of this course, students will be able to:

  1. Explain the biological inspiration for artificial neural networks
  2. Calculate the output of neurons using weighted sums and activation functions
  3. Design appropriate network architectures for classification problems
  4. Trace information flow through forward propagation
  5. Describe how gradient descent enables learning
  6. Diagnose training problems (overfitting, underfitting, learning rate issues)
  7. Evaluate model performance using confusion matrices and backtesting

Course Structure

Part 1: Foundations (Topics 01-04)

Objective: Understand the basic building blocks of neural networks

Topic Title Key Concept
01 Biological Neuron Biological inspiration
02 Single Neuron Weighted sum and activation
03 Problem Visualization Why simple rules fail
04 Neuron Decision Maker Threshold-based decisions

Part 2: Building Blocks (Topics 05-08)

Objective: Master activation functions and understand limitations

Topic Title Key Concept
05 Activation Functions Sigmoid, ReLU, Tanh
06 Linear Limitation XOR problem
07 Sigmoid Saturation Vanishing gradient
08 Boundary Evolution More neurons = better boundaries

Part 3: Network Architecture (Topics 09-12)

Objective: Design and understand multi-layer networks

Forward Propagation

Topic Title Key Concept
09 Network Architecture Layers and parameters
10 Forward Propagation Computing predictions
11 Decision Boundary Linear vs curved separation
12 Feature Hierarchy Layer abstraction

Part 4: Learning Process (Topics 13-16)

Objective: Understand how networks learn from data

Topic Title Key Concept
13 Loss Landscape Error surface
14 Gradient Descent Weight updates
15 Overfitting/Underfitting Training diagnostics
16 Learning Rate Hyperparameter tuning

Part 5: Application (Topics 17-20)

Objective: Apply neural networks to real business problems

Topic Title Key Concept
17 Market Prediction Data Feature engineering
18 Prediction Results Before vs after training
19 Confusion Matrix Performance metrics
20 Trading Backtest Business value

Assessment

Each topic includes 2-3 practice problems with solutions. Students should work through these problems to reinforce understanding.

Self-Assessment Checkpoints:

  • After Part 1: Can you calculate a neuron’s output by hand?
  • After Part 2: Can you explain why XOR requires hidden layers?
  • After Part 3: Can you count parameters in a network architecture?
  • After Part 4: Can you diagnose overfitting from learning curves?
  • After Part 5: Can you interpret a confusion matrix?

Materials

Primary Resource:

  • This GitHub Pages site with all topic pages and exercises

Supplementary:

Contact

This course is part of the Digital-AI-Finance educational initiative.

For questions or feedback, please open an issue on the GitHub repository.

(c) Joerg Osterrieder 2025