Since I didn't notice a git repo, as a means of exploring this idea I opened up antigravity and had Gemini 3.5 Flash do an implementation of these ideas in Julia. I give it a link to this article and to the paper... and 20 minutes later you can take a look here: https://github.com/philtomson/KAN_LUT
Next steps would be to do the FPGA implementation in verilog.
Right. But ... this would limit you to either extremely small models or extremely large FPGA's, yes? If there's a simple machine learning task that requires a sub microsecond latency I can see the point but otherwise??
Yes, definitely: this type of work is applicable in domains where software run on general-purpose processors cannot meet latency or power requirements.
Next steps would be to do the FPGA implementation in verilog.
I've been trying to hit 100,000tokens/s with a 3.28m dumb model, and even this is an order of magnitude too large to benefit.
It appears to be focussed more on latency, than throughput. Happy to be corrected?
One primary application of this work is in high-energy physics (https://home.cern/smarter-decisions-at-the-speed-of-collisio...). Ultrafast and real-time learning is also very applicable for problems in quantum computing, plasma control, etc. (https://arxiv.org/pdf/2602.02005).
p.s. Thanks for posting this and welcome to HN!