Amazon Patents

Computational Systems For Neural Network

Amazon filed these patents around Computational Systems For Neural Network in the last 5 years

# Patent No. Short Description
1. US20200410337A1 Accelerating inferences based on a neural network model using multiple computing engines that perform sub-operations of a neural network operation in parallel
2. US11410639B2 Generating high-quality speech from text using a trained model
3. US10467526B1 Training a neural network model for image similarity analysis using curriculum learning techniques to optimize the selection of training image pairs and the architecture of the model
4. US10522167B1 Using deep neural networks to improve noise cancellation in electronic devices like smartphones
5. US11227014B2 Graph-based convolutional networks (GCNs) that combine elements of a relevant neighborhood and node convolutions to generate enhanced embedding vectors for nodes of a graph
6. US11048979B1 Active learning-based data labeling service that reduces the amount of data to be manually annotated
7. US10860836B1 Generating synthetic image data for training computer vision object detectors using a technique called synthetic composite image data generation
8. US20200380216A1 Evaluating neural network based machine translation models using statistical phrase tables to improve translation quality and reduce resource usage
9. US20200293284A1 Accelerating the quantized computation of an artificial neural network by asymmetrically quantizing floating point data to 8-bit numbers, pre-processing the quantized data to reduce the number of multiplications and summations, and using a simplified matrix multiplication algorithm
10. US20210104245A1 Detecting sentiment in audio data using a trained machine learning model that combines acoustic and lexical features
11. US11501210B1 Accurately inferring or predicting subject matter, fields of interest, or material within various forms of content, such as text, images, videos, or audio
12. US10776626B1 Using artificial intelligence to identify collections of visually complementary items that are aesthetically pleasing when viewed together
13. US10990650B1 Optimizing convolutional neural network (CNN) computations to reduce power consumption and improve performance by avoiding unnecessary computations
14. US10839506B1 Automatically detecting surface flaws in subjects like aerospace components using computer vision techniques
15. US11816446B2 Systolic array architecture that enables efficient processing of neural networks with reduced hardware complexity and power consumption

Download the full patent report containing all of the 241 patents