Data Dialogue

In visual recommendation, for a given product, similar products are retrieved based on their visual similarity rather than their text records. This is particularly useful for classifying products in different subcategories of apparel by their "style" or to customers who want to find products by photos without knowing product names. In this talk, I will show how to build a visual recommender based on triplet neural network using customer browsing data. This is a summer intern project at Target on the data science team.
https://services.math.duke.edu/mcal?abstract-9884