Cristian Daniel Alecsa's Projects
- implementations in PyTorch regarding AAMMSU optimizer
finite difference method for a two-dimensional Poisson BVP with nonconstant spatial coefficients
Monte Carlo based methods for the approximation of given functions. The approximation of PI is also included
- DNN in Keras for Fashion MNIST dataset, including learning rate with decay rate, comparison for dropouts, image prediction, loss and accuracy plots
- DNN from scratch in Python, including different optimizers, cost functions and activation functions
some basic object oriented programming codes made in Python for the comparison between backward ( implicit ) Euler method and the analytical solution of an IVP
- this repository contains some one-dimensional function approximations by neural networks
a generalized version of KMM involving multiple train and test datasets
- includes multi-linear regression on a complete set of features, linear regression made independently on each relevant feature, correlation matrix and feature distributions
This method provides train/test indices according to multiple target columns, and it is suitable for applications involving multiclass-multioutput classification tasks.
an autoencoder method constructed using ensembles of oblique decision trees
a comparison of various optimization algorithms using different objective functions
a comparison of Polyak and Nesterov classical algorithms using convex / nonconvex higher dimensional mappings
- different type of optimizers for 1D and 2D objective functions. A comparison using the energy dissipation, phase portrait, decrease in the objective function.
- descriptive statistics and prediction, using different techniques in Python