Nano Dimension has been granted a patent for a method that trains a new neural network to mimic a target neural network without access to the original training dataset. The method involves probing both networks with input data, detecting differences, and iteratively training the new network to minimize discrepancies. GlobalData’s report on Nano Dimension gives a 360-degree view of the company including its patenting strategy. Buy the report here.

According to GlobalData’s company profile on Nano Dimension, Spacecraft 3D Printing was a key innovation area identified from patents. Nano Dimension's grant share as of April 2024 was 29%. Grant share is based on the ratio of number of grants to total number of patents.

Training new neural network to mimic target neural network

Source: United States Patent and Trademark Office (USPTO). Credit: Nano Dimension Ltd

A recently granted patent (Publication Number: US11961003B2) outlines a method for training a new neural network to mimic a target neural network without direct access to the target network or its original training data. The method involves probing both networks with input data to detect differences in output, generating a divergent probe training dataset based on these differences, and iteratively training the new network to minimize disparities with the target network. This process results in a new neural network with fewer layers and a smaller file size than the original target network. Various techniques are proposed for generating the divergent probe training dataset, including using additional neural networks, evolutionary models, random seed probes, and statistics-based methods. The method also allows for updating the dataset, adding new data, omitting specific data, and removing correlations to tailor the new network's performance.

Furthermore, the patent describes a system and a non-transitory computer-readable medium for implementing this method. The system includes processors configured to train the new neural network, while the computer-readable medium contains instructions for executing the training process. The system and medium store and manipulate the divergent probe training dataset, generate it using different methods, and enable the execution of the trained new neural network in a run-time phase. By utilizing these innovative techniques, the method aims to efficiently train neural networks to mimic target networks while reducing complexity and file size, offering potential applications in various fields requiring neural network training and optimization.

To know more about GlobalData’s detailed insights on Nano Dimension, buy the report here.

Premium Insights


The gold standard of business intelligence.

Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.


GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.

GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.