nicenustian Goto Github PK

bayesian-resnet-convnet-search-regression

bh2igm

This code finds an optimal architecture, search for hyperparameters (using OPTUNA code), trains and make predictions using LYMAN alpha part of simulated 1D super massive black hole spectra and converts it to intergalactic medium gas conditions along the line of sight. However, this code easily be utilized for ANY 1D signals using supervised ML.

extract-small-scale-systems-from-radiative-transfer-cosmological-simulation

Find systems using 4-pixels connectivity in 3d data cube from a radiative transfer simulation

four-params-mcmc

Provides a c++ code to do a fast MCMC search with 4 params quickly. Performs trilinear interpolation saves the chains and check for convergence. However, use has to provide the4d grid and grid of models.

mandelbrot-3d-fractals

Make modified Mandelbrot fractals in 3d by changing the complex number equation and alternating the params. Stack several fractals together to make an environment.

stocks-bayesian-lstm-and-transfer-learning

Predict stock market daily trends with mean one sigma uncertainty, up to several weeks in advance. It combines differents stock from a one sector into one combined model. This serves as a starting point to train each stock again.

super-res-cosmic-images

super-res-hydro

Generate one super resolution one dimensional fields from outputs of Hydrodynamical simulations