17877090. APPARATUS AND METHOD WITH NEURAL NETWORK simplified abstract (Samsung Electronics Co., Ltd.)
Contents
APPARATUS AND METHOD WITH NEURAL NETWORK
Organization Name
Inventor(s)
APPARATUS AND METHOD WITH NEURAL NETWORK - A simplified explanation of the abstract
This abstract first appeared for US patent application 17877090 titled 'APPARATUS AND METHOD WITH NEURAL NETWORK
Simplified Explanation
The patent application describes an apparatus that includes a random-access memory (RAM) with a crossbar array structure. The RAM generates an analog output signal based on an input and the weight of a neural network. The apparatus also includes an analog-to-digital converter (ADC) circuit that generates a digital output signal based on a reference signal and the analog output signal of the RAM. There are two ADC scalers, one for scaling the reference signal and another for scaling the digital output signal.
- The apparatus includes a RAM with a crossbar array structure.
- The RAM generates an analog output signal based on an input and the weight of a neural network.
- An ADC circuit converts the analog output signal to a digital output signal.
- The ADC circuit uses a reference signal for the conversion.
- There are two ADC scalers, one for scaling the reference signal and another for scaling the digital output signal.
Potential Applications
- This technology can be applied in various fields that utilize neural networks, such as artificial intelligence, machine learning, and pattern recognition.
- It can be used in hardware implementations of neural networks, enabling efficient and high-speed processing.
Problems Solved
- The apparatus solves the problem of converting analog signals generated by a RAM to digital signals for further processing.
- It addresses the challenge of scaling the reference signal and the digital output signal to optimize the conversion process.
Benefits
- The use of a crossbar array structure in the RAM allows for efficient and parallel processing of neural network weights.
- The ADC scalers provide flexibility in scaling the reference signal and the digital output signal, allowing for customization and optimization.
- The apparatus enables faster and more accurate processing of neural network inputs, leading to improved performance in applications that rely on neural networks.
Original Abstract Submitted
An apparatus includes: a random-access memory (RAM) configured to generate an analog output signal based on an input and a weight of a neural network, the RAM including a crossbar array structure; an analog-to-digital converter (ADC) circuit configured to generate a digital output signal based on a reference signal and the analog output signal of the RAM; a first ADC scaler configured to scale the reference signal of the ADC circuit; and a second ADC scaler configured to scale the digital output signal generated by the ADC circuit.
