Market Anomaly Detection with Autoencoders - Quant Researcher Interview Question

Problem Description

Yield curves and volatility surfaces often exhibit low-dimensional structures that allow market makers to detect anomalies representing arbitrage opportunities or data errors. Autoencoders leverage this property by learning a compressed representation of the data manifold, where high reconstruction errors indicate significant deviations from normal market behavior. Task Implement a neural network autoencoder from scratch using numpy to detect market anomalies by training on historical yield cur