northlondoner

9 ms·

151.
Gibbs-randomness proposition for energy efficient deep learning(science-memo.blogspot.com)
1 points by northlondoner 1y ago|1 comments
152.
by northlondoner 1y ago
A new technique called DTC can be used to train deep learning with model compression on the fly. This gives rise to massively energy efficient deep learning, reaching almost ~98% reduction in energy use. Moreover, the technique also demons
153.
Differentiating learning curves from Occam curves(science-memo.blogspot.com)
1 points by northlondoner 1y ago|0 comments
154.
In-context learning as a kind of symbolic-AutoML: Lyapunov conjecture for CoTs(hal.science)
2 points by northlondoner 1y ago|0 comments
155.
Have you ever wondered what's the ultimate physical limit of data storage?(science-memo.blogspot.com)
2 points by northlondoner 1y ago|0 comments
156.
LogSoftMax Should Be Preferred?(memosisland.blogspot.com)
3 points by northlondoner 1y ago|0 comments
157.
Lyapunov Conjecture for Chain of Thought(hal.science)
1 points by northlondoner 2y ago|0 comments
158.
Stability in In-Context Learning(dx.doi.org)
1 points by northlondoner 2y ago|0 comments
159.
Stability of In-Context Learning(researchgate.net)
1 points by northlondoner 2y ago|0 comments
160.
It wasn't surprising that NNs got Nobel in Physics(hal.science)
2 points by northlondoner 2y ago|0 comments
161.
by northlondoner 2y ago
Designing in-context data or conversation seems to be a new engineering task.
162.
AutoML: On the fallacy of many-shot in-context learning(memosisland.blogspot.com)
1 points by northlondoner 2y ago|1 comments
163.
Vanilla attention mechanism encodes causality(memosisland.blogspot.com)
2 points by northlondoner 3y ago|0 comments
164.
by northlondoner 3y ago
There are multiple different definitions of Entropy and usage. Boltzmann's and Gibbs's physical entropy is not strictly equivalent to information entropy of Shannon whereby legend says von Neumann told Shannon that call 'info
165.
by northlondoner 4y ago
Well deserved. Impact is immense. Of course, Xerox PARC alumni.
166.
by northlondoner 4y ago
Many congratulations! Glad to hear about UK & EU collaborative innovation in open-source projects. Keep up the fantastic work! Others asked similar question regarding comparable projects. What's your take on OpenFL from Intel? Do y