17808511. REVERSE REINFORCEMENT LEARNING TO TRAIN TRAINING DATA FOR NATURAL LANGUAGE PROCESSING NEURAL NETWORK simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)
REVERSE REINFORCEMENT LEARNING TO TRAIN TRAINING DATA FOR NATURAL LANGUAGE PROCESSING NEURAL NETWORK
Organization Name
INTERNATIONAL BUSINESS MACHINES CORPORATION
Inventor(s)
REVERSE REINFORCEMENT LEARNING TO TRAIN TRAINING DATA FOR NATURAL LANGUAGE PROCESSING NEURAL NETWORK - A simplified explanation of the abstract
This abstract first appeared for US patent application 17808511 titled 'REVERSE REINFORCEMENT LEARNING TO TRAIN TRAINING DATA FOR NATURAL LANGUAGE PROCESSING NEURAL NETWORK
Simplified Explanation
The abstract describes a computer-implemented process for modifying a training dataset using a neural network and reverse reinforcement learning. Here is a simplified explanation of the abstract:
- The process starts by benchmarking the training dataset using a State Of The Art (SOTA) neural network to establish a performance benchmark.
- The training dataset is then divided into smaller slices.
- A selection strategy generator is used to choose a sequence of atomic operations to be applied to one of the slices.
- The selected sequence of operations modifies the slice, creating a revised version.
- Reverse reinforcement learning is performed on the revised slice using the benchmark and the SOTA neural network.
- Finally, the original slice is replaced with the revised slice in the training dataset, resulting in a modified training dataset.
Potential applications of this technology:
- Improving the performance of machine learning models by fine-tuning the training dataset.
- Enhancing the accuracy and efficiency of neural networks in various domains, such as image recognition, natural language processing, and recommendation systems.
Problems solved by this technology:
- Addressing the limitations of traditional training dataset modification methods by using a combination of benchmarking, atomic operations, and reverse reinforcement learning.
- Overcoming the challenges of manually modifying large training datasets by automating the process.
Benefits of this technology:
- Increased accuracy and performance of machine learning models by optimizing the training dataset.
- Time and cost savings by automating the dataset modification process.
- Improved generalization and robustness of neural networks by incorporating reverse reinforcement learning.
Original Abstract Submitted
A computer-implemented process for modifying a training dataset includes the following operations. The training dataset is benchmarked using a State Of The Art (SOTA) neural network to determine a benchmark for the training dataset. The training set is divided into a plurality of slices. A sequence of a plurality of atomic operations are selected using a selection strategy generator operating on one of the plurality of slices. The sequence of the plurality of atomic operations is applied to modify the one of the plurality of slices to generate a revised one of the plurality of slices. Reverse reinforcement learning is performed on the revised one of the plurality of slices using the benchmark and the SOTA neural network. The training dataset is modified by replacing the one of the plurality of slices with the revised one of the plurality of slices to generate a modified training dataset.
