Agustinus Kristiadi

generative-models. Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow.

MCMC. Collection of Monte Carlo (MC) and Markov Chain Monte Carlo (MCMC) algorithms applied on simple examples.

hipsternet. All the hipster things in Neural Net in a single repo

probabilistic-models. Collection of probabilistic models and inference algorithms

controlled-text-generation. Reproducing Hu, et. al., ICML 2017's "Toward Controlled Generation of Text"

natural-gradients. Collection of algorithms for approximating Fisher Information Matrix for Natural Gradient (and second order method in general)

last_layer_laplace. Last-layer Laplace approximation code examples

pub-ready-plots. A very opiniated publication-ready plotting library

cnn-levelset. Source code for Kristiadi and Pranowo, 2017's "Deep Convolutional Level Set Method for Image Segmentation"

wiseodd.github.io. Agustinus‘ personal website and blog

lapeft-bayesopt. Discrete Bayesian optimization with LLMs, PEFT finetuning methods, and the Laplace approximation.

lula. Companion code for the paper "Learnable Uncertainty under Laplace Approximations" (UAI 2021).

compound-density-networks. Implementation of: Kristiadi, Agustinus, and Asja Fischer. "Predictive Uncertainty Quantification with Compound Density Networks." (2019).

cuda-pso. Comparison between CUDA and CPU for PSO using 3-dimensional Levy function as test case

laplace-bayesopt. Laplace approximated BNN surrogate for BoTorch

bayesian_ood_training. Jupyter Notebook

swarm-image-segmenter. Image segmentation using swarm intelligence, accelerated using GPGPU (CUDA)

llm-bayesopt-exps. Official experiment code for the "Sober Look at LLMs for Material Discovery" paper.

phd_thesis_template. TeX

rgpr. Companion code for the paper "An Infinite-Feature Extension for Bayesian ReLU Nets That Fixes Their Asymptotic Overconfidence" (NeurIPS 2021)

two-funds-rebalancer. Given your current portfolio value, desired allocation after rebalancing, and the amount of cash you have, this script will output how much of your cash should be used to buy stock/bonds (in percent). Only works for portfolios with two funds, e.g. the Couch Potato and classic two-funds 60-40 portfolios.

ziglox. Lox compiler written in Zig

wiseodd.

nanoGPT. The simplest, fastest repository for training/finetuning medium-sized GPTs.

bo-async-feedback. Python

fast-neural-style. pytorch implementation of fast-neural-style

aoc-2024. Agustinus' Advent of Code 2024 Solutions in Gleam

MA-INF-4222-NLP-Lab. MA-INF 4222: NLP Lab (University of Bonn)

higher_order_invariance. Code for "Accelerating Natural Gradient with Higher-Order Invariance"