Increase accuracy and reduce network size, easy as PAI!
Welcome to Perforated A.I.
Perforated AI is a start-up company founded by Dr. Rorry Brenner in 2022. Rorry earned his Ph.D in Neuroscience from USC and B.S. in Computer Science from Carnegie Mellon. His experience includes 8 years of deep learning work with 4 years in industry on commercial and government projects. Perforated A.I. specializes in the implementation of a patent protected modification to the backpropagation algorithm. This biological neuroscience based modification can quickly be added to existing deep learning systems to reduce error and network size.
If you're interested in reducing your network sizes and increasing accuracy check out our current test results and reach out with any questions!
Company Summary
Disruptive Technology in Deep Learning
Perforated Backpropagation is the first major change to the underlying backpropagation paradigm
Potential Partners
Machine learning groups in academia and industry who use PyTorch
Methodology
Neuroscience-based learning paradigm within modern deep learning architectures
It can be added to any established PyTorch model, often with under an hour of coding
Improves accuracy and network size
Tested with convolutional ResNets, LSTM time series networks, transformers, and more
Adding Perforated Backpropagation to state-of-the-art networks
AUC on Tox 21 dataset 41% closer to a perfect score
Validation error on SUPERB dataset decreased by 12%
Reduction in network size on Physionet Dataset to 10.7% initial size while maintaining AUC score
Goal
Finding groups interested in trying a demo
Integrate into your existing system
Evaluate our algorithm on a real world dataset and real world application
Improve your accuracy and reduce your network size
Current Status
Minimum viable product complete and ready for use