Stanislav Fort

exploring_the_limits_of_OOD_detection. Code to replicate the key results from Exploring the Limits of Out-of-Distribution Detection (https://arxiv.org/abs/2106.03004) by Stanislav Fort, Jie Ren, Balaji Lakshminarayanan, published at NeurIPS 2021.

mythos-jagged-frontier.

gaussian-prototypical-networks. Gaussian prototypical network architecture for few-shot learning

Direct_Ascent_Synthesis. A demo for the Direct Ascent Synthesis: Hidden Generative Capabilities in Discriminative Models paper (https://arxiv.org/abs/2502.07753)

dissect-git-re-basin. Replicating and dissecting the git-re-basin project in one-click-replication Colabs

OpenAI_CLIP_adversarial_examples. Developing adversarial examples and showing their semantic generalization for the OpenAI CLIP model (https://github.com/openai/CLIP)

ensemble-everything-everywhere. The code for the Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness paper

adversaries_to_convnext. Adversarial examples to the new ConvNeXt architecture

adversaries_to_OOD_detection. Jupyter Notebook

multi-attacks. Replicating the basic adversarial multi-attack experimental results

ulam-spiral. The Ulam prime spiral generator

hilbert-curves. Generating 3D Hilbert curves

slice-dice-optimize. Code to replicate the key findings of the paper /What does a deep neural network confidently perceive? The effective dimension of high certainty class manifolds and their low confidence boundaries/

stanislavfort.github.io. A research blog on machine learning, artificial intelligence and physics

singular-value-modified-training-for-fun. Training DNNs on data with singular values removed / kept, testing on similarly modified test sets

Data-Visualization. A collection of tips and tricks for data visualization in Python.

ensemble-landscape-cuts. Reproducing plots and loss landscape cuts from the paper *Deep Ensembles: A Loss Landscape Perspective* (https://arxiv.org/abs/1912.02757)

markov_language_model. A character-level probabilistic language generator based on Markov chains

GettingStarted. ...with git and GitHub

initialization-optimum-connectivity-experiments. Assorted experiments to verify the connectivity of DNN initializations to optima in the weight space

classical_implementation_grovers_algorithm. A classical (non-quantum) implementation of the Grover's algorithm (quantum)

ensemble-everything. A website for the academic paper Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness

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

fast_template. A template for really easy blogging with GitHub Pages

vision_transformer. Jupyter Notebook

deepPylos. Deep reinforcement learning for the board game Pylos