18068987. EFFICIENT MACHINE LEARNING MODEL ARCHITECTURES FOR TRAINING AND INFERENCE simplified abstract (QUALCOMM Incorporated)

From WikiPatents
Jump to navigation Jump to search

EFFICIENT MACHINE LEARNING MODEL ARCHITECTURES FOR TRAINING AND INFERENCE

Organization Name

QUALCOMM Incorporated

Inventor(s)

Joseph Binamira Soriaga of San Diego CA (US)

Hossein Hosseini of San Diego CA (US)

EFFICIENT MACHINE LEARNING MODEL ARCHITECTURES FOR TRAINING AND INFERENCE - A simplified explanation of the abstract

This abstract first appeared for US patent application 18068987 titled 'EFFICIENT MACHINE LEARNING MODEL ARCHITECTURES FOR TRAINING AND INFERENCE

The present disclosure introduces techniques for enhancing machine learning processes through the use of data tensors in neural networks.

  • A data tensor is produced as output from a neural network layer.
  • The data tensor is split into a first subset and a second subset using a tensor splitting operation.
  • The second subset is stored for later use, while the first subset is passed on to the next layer of the neural network.
  • Parameters of the neural network layer are adjusted based on the stored second subset of the data tensor.

Potential Applications: - This innovation can be applied in various fields such as image recognition, natural language processing, and predictive analytics. - It can improve the accuracy and efficiency of machine learning models in tasks like classification and regression.

Problems Solved: - Enhances the performance of neural networks by refining parameters based on stored data subsets. - Facilitates better learning and decision-making processes in complex data environments.

Benefits: - Increases the effectiveness of machine learning algorithms. - Enables more accurate predictions and classifications in various applications. - Enhances the overall performance and efficiency of neural networks.

Commercial Applications: "Optimizing Machine Learning Processes with Data Tensors in Neural Networks" This technology can be utilized in industries such as healthcare, finance, and e-commerce for tasks like medical diagnosis, fraud detection, and personalized recommendations. It has the potential to revolutionize how businesses leverage data for decision-making and automation.

Questions about the technology: 1. How does the use of data tensors improve the performance of neural networks? 2. What are the key advantages of refining parameters based on stored data subsets in machine learning models?


Original Abstract Submitted

Certain aspects of the present disclosure provide techniques for improved machine learning. A data tensor is generated as output from a layer of a neural network. A first subset of the first data tensor and a second subset of the first data tensor are generated using a tensor splitting operation. The second subset of the first data tensor is stored, and the first subset of the first data tensor is provided to a subsequent layer of the neural network. One or more parameters of the layer of the neural network are refined based at least in part on the stored second subset of the first data tensor.

Retrieved from "http://wikipatents.org/index.php?title=18068987._EFFICIENT_MACHINE_LEARNING_MODEL_ARCHITECTURES_FOR_TRAINING_AND_INFERENCE_simplified_abstract_(QUALCOMM_Incorporated)&oldid=77047"