SOFTWARE-BASED MASS CUSTOMIZATION OF ARTIFICIAL NEURAL NETWORKS AND ITS BENEFITS (RFT-681)

Invention Summary

Artificial intelligence (AI) and machine learning (ML) technologies are transforming various industries by enabling sophisticated data analysis and decision-making. The adaption of these technologies is often hindered by complexity, the lack of explainability, and resource requirements. This technology supports providing a user-friendly software platform designed for the mass customization of artificial neural networks (ANNs). The platform employs a diagramming interface, allowing users to create, visualize, and customize neural network models with ease.

computer monitors imaging neural networks| NDSU Research Foundation

Key Features

  • Enables a Drag-and-Drop Interface: users can construct neural networks by dragging and dropping neuron and link objects, streamlining the model creation process
  • Batch Creation: allows the generation of multiple neurons and layers simultaneously, simplifying the development of complex models
  • Neuron-Level Customization: users can modify individual neurons' activation functions, weight initializations, and connectivity, facilitating the creation of heterogeneous neural networks
  • Coordinate Mapping System: a Cartesian plane coordinate system helps users keep track of neural network components, improving model management

Benefits

  • Accessibility: simplifies the process of building and customizing neural networks, making advanced AI and deep learning (DL) technologies accessible to users with varying levels of expertise
  • Flexibility: enables detailed customization at the neuron level, allowing users to tailor models to specific applications and experiment with different configurations
  • Efficiency: automates the creation of backend logic, saving time and reducing the need for extensive technical knowledge
  • Scalability: supports the creation of large-scale, complex models through batch creation and easy visualization tools

Applications

  • Healthcare: developing diagnostic models and personalized treatment plans
  • Finance: fraud detection, risk assessment, and predictive analytics
  • Autonomous Vehicles: enhancing decision-making processes for vehicle navigation and safety systems
  • Cybersecurity: intrusion detection and threat analysis
  • Natural Language Processing: improving language translation, sentiment analysis, and conversational AI

Patent

This technology has a U.S. Patent Pending and is available for licensing/partnering opportunities.

Contact

NDSU Research Foundation
info(at)ndsurf(dot)org
(701) 231-8173

NDSURF Tech Key

RFT, 681, RFT681

Inquire about this technology >