The repository contains research and educational projects.
- Our site
- Department site (Russian version)
- Hugging face
- Mail: research.n.math(at)gmail.com
Before creating new repository, please read this friendly manual
The repository contains research and educational projects.
Before creating new repository, please read this friendly manual
Intelligent Systems course on Deep Learning
Jupyter Notebook 31
Implementation of Implicit Reparameterization Trick
[BMM 24-25] HippoTrainer: Gradient-Based Hyperparameter Optimization
Jupyter Notebook 12
BMM: deep learning model complexity estimation. Оценка сложности глубоких моделей.
Removing groups of weights from a neural network based on the deep learning computation graph analyzing and estimation of information flow transferred through it.
Метод, реализующий и сравнивающий методы оценки градиентов с низкой дисперсией для обучения моделей со смесями распределений.
[TODO] By inductive bias, we mean the set of assumptions that a model uses to make predictions on unseen data. Not all model architectures possess the necessary biases. We consider a set of models with inductive biases induced by different training datasets. We propose a method to transfer these biases to a model that lacks them.