3 ms·
Thank you! Definitely a group effort. Two big things that contributed to the choice: We knew we didn’t want the system to have to store memory on each edge so t
by RaisinLoaf69 5y ago
Thank you! Definitely a group effort. Two big things that contributed to the choice: We knew we didn’t want the system to have to store memory on each edge so that eliminated a lot of our options. In our experimental learning rule we only had to compare voltages which is actually easier to do if we have two simultaneous networks.
- mlajtos 5y agoInteresting, I haven't seen this twin approach anywhere in ANNs. (I know about Barlow twins and Siamese nets, but this is different.) How did you decide on the topology of the network/graph?
- RaisinLoaf69 5y agoSince we have no global processor, each edge is changing using only local information, a lot of stuff that makes sense in our network doesn't really make sense in ANNs and vice versa. For example, in our newest paper (https://arxiv.org/abs/2201.04626 https://arxiv.org/abs/2201.04626) we desynchronize the updates of our edges. Instead of changing the entire system all at once, we change random parts of it each training step. This doesn't really make sense to do in an ANN where you require global information for every edge update. The shape of the network is actually inspired by jamming solids (we're a soft matter lab), but is completely arbitrary. We've done a ton of different shapes and sizes in simulation.