MEA-Based Biological Neural Network Processing for Lower-Power AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for artificial neural networks (ANNs) in various applications is accompanied by significant time, energy, and data requirements for training, which existing technologies struggle to efficiently address.
Innovation Solution
A biological and artificial neural network (BANN) system is employed, comprising a multi-electrode array (MEA) with biological neural networks (BNNs) and trained statistical models, which encodes input signals as stimulation patterns, measures responses, and processes them through ANNs to perform tasks efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional artificial neural networks are used to process complex tasks, then computational capability is improved, but power consumption and training time increase significantly
Solution Approach 1:
The patent introduces biological neural networks as an intermediary component between input data and artificial neural network processing. The BNN performs preliminary processing of time-series data, extracting features and reducing data complexity before passing to the ANN, thereby reducing the computational burden and power consumption of the ANN while maintaining computational capability.
Solution Approach 2:
The patent segments the neural processing task into two distinct parts: biological neural network processing for initial data interpretation and feature extraction, followed by artificial neural network processing for higher-level computation. This segmentation allows each component to operate more efficiently, with the BNN handling noisy, real-time data and the ANN focusing on refined computational tasks.
2Measurement precision
If traditional artificial neural networks are used for complex tasks, then accuracy is improved, but training iterations and time requirements increase
Solution Approach 1:
The biological neural network performs preliminary action by pre-processing input data and extracting meaningful features before the artificial neural network begins its training process. This preliminary processing reduces the dimensionality and noise in the data, allowing the ANN to converge faster during training while achieving the same or better accuracy.
Solution Approach 2:
The system implements feedback mechanisms where the BNN continuously monitors input data quality and adjusts its processing accordingly, providing optimized input to the ANN. This feedback loop enables the system to adapt to varying data conditions, maintaining high accuracy while reducing the number of training iterations needed.
3Use of energy by moving object
If biological neural networks are used alone, then energy efficiency is improved, but computational precision and reliability decrease
Solution Approach 1:
The patent merges biological neural networks and artificial neural networks into a hybrid BANN system that combines the energy efficiency of BNNs with the computational precision of ANNs. The BNN handles energy-efficient preliminary processing while the ANN provides reliable, precise computation for final results, achieving both energy efficiency and high reliability simultaneously.
Solution Approach 2:
Different parts of the hybrid system have specialized functions optimized for their strengths: the BNN portion is optimized for energy efficiency and real-time processing of noisy data, while the ANN portion is optimized for computational precision and reliability. This local quality differentiation allows each component to excel at its specific task while contributing to overall system performance.
Data Source
AI summary
Techniques for using a biological and artificial neural network (BANN) system to perform a task. The BANN system comprises a multi-electrode array (MEA); a biological neural network (BNN) comprising neurons arranged on the MEA, a trained statistical model trained using inputs generated using responses of the BNN to training data inputs; and at least one processor. The method comprises using the BANN system to receive an input signal; encode the input signal to generate a stimulation pattern; stimulate the BNN by using the MEA to generate electrical signals in accordance with the stimulation pattern; measure, using the MEA, a response of the BNN responsive to the stimulating; generate, based on the measured response, an input for the ANN; process the input with the trained statistical model to obtain corresponding output; and use the output from the trained statistical model in furtherance of performing the task.