3d Generative modeling

Diffusion nets were great success with image generation. I'll make current This overview will help you to formS

Prerequisites:

• duality between wave and particle

Main thesis: neural net can be viewed as freezing. Under very low temperature matter changes and quantum effect appears.

Rigid body theory

Rigid body is built from latices

Rigid body lattice

Points of lattice are field with positive atom kernel and are covered with dust of electrons

Phonon is representation of lattice perturbation. They helps to build mechanical intuition of phase transitions.

Boson and fermi statistics

\[\]

Quantum physics swept away not

Most notoriously for statistical description is that. Therefore it’s permuttative. That brings insights about energy distribution of particles/

Quasi Particles

Bose einstein condesat

All particles are placed at lowest energy level with gap to another big enough

\[\]

Superfluid is matter that doesn’t have vicosity. It means it doesn’t have friction with surface along with it’s traveling.

Lattice can vibrate

Actually, that wasn’t true. It happens that there is special speed called Landau speed under which rotons start to

In very low temperatures.

Most essential is a gap between level of excitation.

Superfluidity comes due to fact that quasiparticles doesn’t have enough

Fermi statistics

£

Explanation of superconductivity

https://en.wikipedia.org/wiki/Bogoliubov_transformation

Сurrent in conductors is built from unpaired electrons, that travels freely through lattice of matter. Stream of such electrons builds current

Current

Resistance of conductor comes from fact that lattice starts to vibrate under stream. Such vibrations can represented as phonons, which

Literally, their interaction can be rewriten in motion mechanics maneer

\[k + k' = k\\ E + E = E\]

Electron scattering in flow results in resistance. You can even achieve exact formula.

Key insights:

• if something has resistance it means there are some mechanism of scattering of stream

Let’s try to apply to neural net.

Resistance here is connected with fact that nets need to learn

Local minimas - are scattering and can be replaced with “gradions” - quasiparticles that scatters gradient descent.s

Gradions

It explains why material current resistance

Some formulas:

Training is performed in job, that is called in CI script. Than model proceeds to designated model bucket

Yet electrons in streams meet other particles called phonons. It’s nature comes from reaction of lattice due to distortion.

Small distortion builds harmonic occillations. | | |:–:| | Training is performed in job, that is called in CI script. Than model proceeds to designated model bucket |

Fewer intuition

Suc

Training is performed in job, that is called in CI script. Than model proceeds to designated model bucket

\[\]

#

Statistics

Different perspective.

\[L = - y log p - (1-y) log p\]

Something small is something big

Small changes in form of meandr brings

Current

That it highly used in guitar amplificators for bringer richer harmonics to sound