categoricalcrossentropy Goto Github PK

adaptive_pooling

The input to the algorithm is the original image, the coordinates of the foci, the percentage of pixels without sub-sampling, and the compression ratio of the rest of the image. Next, the algorithm pools the image in layers: the strongest compression covers the entire area of the image, and part of the image within the focus remains unprocessed. As a result of the algorithm, a certain number of matrices is obtained, equal to the number of pooling layers, these matrices are rectified into a vector with fewer parameters compared to the original image.

brain

A model of neurons learning according to the Hebb rule.

covid-charts

Automatic update of the covid-19 dataset and displaying graphs

filter_visualization

gravity

Simulation of gravitational interaction

neuroevolution

Neuroevolution is a set of algorithms that use artificial evolution to train a neural network. This learning mechanism has a number of advantages over classical methods of model optimization: neuroevolution algorithms do not require calculating the gradient of the loss function; the network architecture can automatically change and adapt to a specific task. It is also effective in reinforcement learning tasks.

simple_telegram_bot_books

Код бота с первой практики

telegram-history-bot

telegram_bot_for_test

Код бота со второй практики

vae

An autoencoder is a combination of two neural networks: an encoder and a decoder. The encoder learns to "compress" input data into a small vector of numbers, highlighting the most important features. The decoder restores the original data from the received features. The autoencoder helps to get a high-level representation of data, which can be used in processing and generating a large amount of information.

vk_bot_test