18596535. USING EMBEDDING FUNCTIONS WITH A DEEP NETWORK simplified abstract (Google LLC)

USING EMBEDDING FUNCTIONS WITH A DEEP NETWORK

Organization Name

Google LLC

Inventor(s)

Gregory S. Corrado of San Francisco CA (US)

Kai Chen of San Bruno CA (US)

Jeffrey A. Dean of Palo Alto CA (US)

Gary R. Holt of Murrysville PA (US)

Julian P. Grady of Pittsburgh PA (US)

Sharat Chikkerur of Pittsburgh PA (US)

David W. Sculley, Ii of Pittsburgh PA (US)

USING EMBEDDING FUNCTIONS WITH A DEEP NETWORK - A simplified explanation of the abstract

This abstract first appeared for US patent application 18596535 titled 'USING EMBEDDING FUNCTIONS WITH A DEEP NETWORK

Simplified Explanation: The patent application describes methods, systems, and apparatus for using embedded functions with a deep network to process input data and predict labels.

• **Key Features and Innovation:**

   - Utilizes embedding functions for different feature types independently.
   - Processes numeric values using a deep network for alternative representation.
   - Uses logistic regression classifier for label prediction.

• **Potential Applications:**

   - Natural language processing
   - Image recognition
   - Sentiment analysis

• **Problems Solved:**

   - Efficient processing of diverse feature types
   - Improved accuracy in label prediction

• **Benefits:**

   - Enhanced performance in machine learning tasks
   - Scalability for large datasets
   - Versatility in handling various types of data

• **Commercial Applications:**

   - AI-powered recommendation systems
   - Fraud detection algorithms
   - Personalized marketing strategies

• **Prior Art:**

   - Researchers in the field of deep learning and machine learning
   - Academic papers on embedding functions and deep networks

• **Frequently Updated Research:**

   - Latest advancements in deep learning models
   - New techniques for feature embedding

Questions about the Technology: 1. What are the potential limitations of using embedded functions with a deep network? 2. How does the logistic regression classifier improve label prediction accuracy?

Original Abstract Submitted

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using embedded function with a deep network. One of the methods includes receiving an input comprising a plurality of features, wherein each of the features is of a different feature type; processing each of the features using a respective embedding function to generate one or more numeric values, wherein each of the embedding functions operates independently of each other embedding function, and wherein each of the embedding functions is used for features of a respective feature type; processing the numeric values using a deep network to generate a first alternative representation of the input, wherein the deep network is a machine learning model composed of a plurality of levels of non-linear operations; and processing the first alternative representation of the input using a logistic regression classifier to predict a label for the input.