Wells Fargo AI leader Freddy Lecue describes training a transformerA transformer is a neural network architecture that uses attention to model relationships between elements of its input. on sequences of customer transactions and adapting the representationA latent representation is an internal numerical encoding an AI model uses to represent and process information. for fraud detection. He explains why tokenizing tabular features, including time, amounts and categories, required particular care.
Lecue reports that the model improved fraud detection and reduced false positivesAn AI evaluation metric is a defined measurement used to quantify a particular aspect of an AI system's performance. relative to the bank’s XGBoost baseline, while meeting latency requirements. He says other applications are under investigation. These are the speaker’s reported results, not an independent assessment.
Watch on YouTube
