A new paradigm of machine learning.
Train models faster, with less data, on CPU.
For most applications, RCL outpaces neural networks by orders of magnitude in training and inference.
For every application thus far, RCL has outperformed state of the art.
RCL trains on CPU, instead of specialized hardware (GPU, TPU, NPU), and requires less compute.
Randomness as a filter.
RCL makes signal strength a user-adjustable parameter.
Less data needed.
RCL’s filter illuminates patterns as soon as they become useful.
RCL can be applied to data of all kinds. RCL-T is just the beginning.
Random Contrast Learning (RCL) is a new style of machine learning invented in 2022 by Dr. Morten Middelfart, Sam Martin, and Ben Martin, who drew upon:
- 7 years of development by Lumina;
- 40 years of pioneering work in the use of randomness in computer science; and
- Engagement with the philosophical tradition of phenomenology.
Train. Test. Repeat.
Time, cost of compute, and cost of specialized hardware often prohibit the iterated training of neural networks. The speed and accessibility of RCL make experimentation and rapid refinement possible.
Iterate without limitations.
Let’s build the future together.
Lumina hosts a showcase where users can share access to their applications with the community. Utilize inference to interact with hosted models.