One of the most exciting and seemingly ubiquitous recent topics in deep learning (DL) is without a doubt multi-task learning. Multi-task learning (MTL) is a model training technique where you train a single deep neural network on multiple tasks at the same time. Though it may seem a little counter-intuitive…

Teaching a machine to categorize something into multiple, non-exclusive groups can feel a little lot juggling. You’ve got to keep track of a lot of things and if any one thing gets out of balance, everything can come crashing down. Here at GumGum, we use machine learning and AI to…

At GumGum, we train machine learning models on unstructured data, like text and images. An integral part of model development is model evaluation. Especially for classification tasks, there are several metrics out there. Most known are accuracy, precision, recall, the F-beta score, and the ROC AUC. Each of those gives…

At GumGum, we train machine learning models on unstructured data, like text and images. An integral part of model development is model evaluation. Especially for classification tasks, there are several metrics out there. Most known are accuracy, precision, recall, the F-beta score, and the ROC AUC. Each of those gives…

At GumGum, we train machine learning models on unstructured data, like text and images. An integral part of model development is model evaluation. Especially for classification tasks, there are several metrics out there. Most known are accuracy, precision, recall, the F-beta score, and the ROC AUC. Each of those gives…

Splitting multi-label data in a balanced manner is a non-trivial task which has some subtle complexities. In this blog post, I review several algorithm implementations and attempt to find the best solution for the data that I’ve been working with. Overview: One reoccurring problem for applied scientists on the GumGum Verity…

“With great models comes slower inference speeds”. Deep Learning has evolved immensely and it has Transforme(r)d NLP completely in the past 5 years. Although these models do achieve state of the art results on various NLP tasks, the models are really big and slow. Bringing these models to production…

A complete demo for developing locally and deploying on Databricks — Real-time machine learning inference at scale has become an essential part of modern applications. GumGum’s Verity engine powers the industry’s most sophisticated contextual targeting product by analyzing thousands of digital content every second around the clock. …