Is it right to basically juxtapose #deepNetworks against #knowledgeBasedReasoning ?
Eg. Hopfield nets versus case based reasoning.
Where deep networks generally simulate parts of biological mechanisms, often on very fast/hungry hardware very quickly, often involving Hidden Layers for what's-actually-being-done.
Whereas in knowledge based reasoning, high level descriptions of the world are being reasoned about eg as cases, (+ situation calculus say). Decision trees.
Efficient optimization of deep neural quantum states toward machine precision
Neural quantum states have emerged as a novel promising numerical method to solve the quantum many-body problem. However, it has remained a key challenge to train modern large-scale deep network architectures, which would be vital in utilizing the full power of the underlying artificial neural networks. In this recent preprint we take on this challenge:
Neural quantum states (NQSs) have emerged as a novel promising numerical method to solve the quantum many-body problem. However, it has remained a central challenge to train modern large-scale deep network architectures to desired quantum state accuracy, which would be vital in utilizing the full power of NQSs and making them competitive or superior to conventional numerical approaches. Here, we propose a minimum-step stochastic reconfiguration (MinSR) method that reduces the optimization complexity by orders of magnitude while keeping similar accuracy as compared to conventional stochastic reconfiguration. MinSR allows for accurate training on unprecedentedly deep NQS with up to 64 layers and more than $10^5$ parameters in the spin-1/2 Heisenberg $J_1$-$J_2$ models on the square lattice. We find that this approach yields better variational energies as compared to existing numerical results and we further observe that the accuracy of our ground state calculations approaches different levels of machine precision on modern GPU and TPU hardware. The MinSR method opens up the potential to make NQS superior as compared to conventional computational methods with the capability to address yet inaccessible regimes for two-dimensional quantum matter in the future.
Hey, can Mastodon and Sigmoid help me find papers/insights?
I am looking for papers that examine kernels in CNNs, specifically the differences between lower and higher layer kernels, better if of the same dimensions.
Are 3x3 kernels that many, that significantly different? Are they doing the same things, are they processing textures to shape, from layer to layer?
#CommonLisp #Gopher #binryhop #deepNetworks #asdf #lisp
gopher://tilde.institute/1/~screwtape/binry-hop/
https://gopher.floodgap.com/gopher/gw.lite?=tilde.institute+70+312f7e7363726577746170652f62696e72792d686f702f
gopher://gopher.club/1/users/screwtape/
https://gopher.floodgap.com/gopher/gw.lite?=gopher.club+70+312f75736572732f7363726577746170652f
I was redeveloping my nascent binry-hop deep hopfield network package to use package-inferred-system, so different sorts of components and data (book)? can be cooked into one overarching system but loaded separately.
I'm happy with it; and I believe in using asdf strongly idiomatically. Commentary?
vintage screwlisp account
Markus Heyl
Fabrice Tshimanga
heise online (inoffiziell)