Block Drop

A platform for both immense data analysis and immutable data storage and management.

deco-blob-1 decoration
graphical divider

 

Block Drop technology was acquired by UpLinkd Group and integrated into our LINKD System to provide for both immense data analysis and immutable data storage and management.

Key Features & Benefits

Residual Networks

Dropping & keeping residual blocks in a pre-trained ResNet based on a policy network's output.

Classify/Detect/Segment

We drop blocks as a way of reducing the number of parameters required for training our ML models.

Immutable Data Storage

Block Drop allows for tamper-proof, auditable data records that cannot be altered or deleted.

Pattern Analysis

With Block Drop & LINKD companies can identify early trends and patterns within transaction data.

Block Drop is an innovative technology developed post-acquisition by UpLinkd Group

This technology enables the dropping and keeping of residual blocks in a pre-trained ResNet based on a policy network's output. The active blocks are then evaluated to make a prediction.

Block Drop technology has been implemented in various applications, including:

  • image classification
  • object detection
  • segmentation

It has been shown to improve accuracy and reduce the number of parameters required for training. This technology has the potential to revolutionize the field of deep learning by enabling the creation of more efficient and accurate models.

Block Drop technology also has applications in data storage. We have developed an immutable data storage system to allow for the creation of tamper-proof, auditable data records that cannot be altered or deleted.

“We have developed an immutable data storage system to allow for the creation of tamper-proof, auditable data records that cannot be altered or deleted.”
– CEO, UpLinkd Group

With Block Drop technology, UpLinkd Group has created a powerful tool for improving the efficiency and security of deep learning and data storage.

By reducing computational costs and memory usage while improving accuracy and security, this technology has the potential to transform the way we approach these fields.