Free module · Machine learning & adaptive decisions
Deep Learning: Sequences, Representations & Financial Decisions
Understand the operations inside the network, then test whether added capacity earns its complexity.
backpropagation · causal CNNs · LSTM · attention · fine-tuning · text
The building blocks
A neural network is a composition of small adjustable functions. Understand one weighted sum, one nonlinear function and one error update before adding layers or attention.
- Follow one neuron and its error
- Use the chain rule to change a weight
- Add sequence structure only when the task needs it
Lessons in this module
- Neural networks and backpropagation from first principles
- Causal windows, temporal convolutions and order-book tensors
- Recurrent networks and LSTM gates
- Attention and transformers: which history can the model use?
- Forecast loss versus trading loss, turnover and differentiable decisions
- Fine-tuning, financial text and foundation-model contamination
Practice and apply
- Take a gradient step — x=2, w=0, b=0, y=1, tanh neuron, learning rate .1.
- Compute a causal filter — Current and two prior values: 4,2,1. Weights: .5,.3,.2.
Work through the practice exercises · Quant development tools