17525999. PARAMETER REDUNDANCY REDUCTION METHOD simplified abstract (International Business Machines Corporation)
Contents
PARAMETER REDUNDANCY REDUCTION METHOD
Organization Name
International Business Machines Corporation
Inventor(s)
PARAMETER REDUNDANCY REDUCTION METHOD - A simplified explanation of the abstract
This abstract first appeared for US patent application 17525999 titled 'PARAMETER REDUNDANCY REDUCTION METHOD
Simplified Explanation
The patent application describes a method, computer program product, and computer system for generating and using a basic state layer. This involves training multiple task models on a pre-trained backbone model, where each task model has different parameter matrices in its feature layers. An encoder-decoder model is then trained, which compresses the parameter matrices into hidden layers to generate a basic state model. The decoder part of the model generates an output layer identical to the input layer.
- The patent application describes a method for generating and using a basic state layer.
- Multiple task models are trained on a pre-trained backbone model.
- Each task model has different parameter matrices in its feature layers.
- An encoder-decoder model is trained, which compresses the parameter matrices into hidden layers.
- The encoder part of the model maps and compresses the parameter matrices into hidden layers to generate a basic state model.
- The decoder part of the model generates an output layer identical to the input layer.
Potential Applications
- This technology can be applied in various fields where multiple task models need to be trained and used.
- It can be used in machine learning applications for tasks such as image recognition, natural language processing, and speech recognition.
- The basic state layer generated by this method can be used as a representation of the input data for further processing or analysis.
Problems Solved
- This technology solves the problem of training multiple task models with different parameter matrices.
- It provides a method to compress and represent the parameter matrices into a basic state model.
- The use of an encoder-decoder model allows for efficient mapping and generation of the output layer.
Benefits
- The method allows for efficient training and utilization of multiple task models.
- It provides a compressed representation of the input data through the basic state model.
- The use of an encoder-decoder model enables the generation of an output layer identical to the input layer.
Original Abstract Submitted
A method, computer program product, and computer system for generating and using a basic state layer. N task models are provided (N ≥ 2). Each task model was trained on a same pre-trained backbone model. Each task model includes M feature layers and a task layer (M ≥ 1). Each feature layer of each task model includes a parameter matrix that is different for the different models. An encoder-decoder model is trained. The encoder-decoder model includes sequentially: an input layer, an encoder, M hidden layers, a decoder, and an output layer. The encoder is a neural network that maps and compresses the parameter matrices in the input layer into the M hidden layers, which generates a basic state model. The decoder is a neural network that receives the basic state model as input and generates the output layer to be identical to the input layer.
