non-financial development institution
major organizer of congress and exhibition events
RC personal account
Восстановление пароля
Введите адрес электронной почты или телефон, указанные при регистрации. Вам будет отправлена инструкция по восстановлению пароля.
Некорректный формат электронной почты или телефона
22 July 2019
Vitaliy Milke

Vitaliy Milke on Pre-trained Neural Networks in Medicine and Genetics

The first algorithm for training neural networks was formulated by the Canadian neuropsychologist Donald O. Hebb, who studied neuron interaction and was interested in the principle governing their combination into ensembles. Hebb’s ideas were initially criticized, though some years later a group of American scientists were able to model an artificial neural network that could distinguish squares from other geometric shapes.

In 2014, a revolutionary event took place: Deep Artificial Neural Networks were able to recognize objects in pictures with an accuracy of more than 95%, exceeding human capabilities. From this moment, neural networks began their triumphal procession across the globe.

Neural networks of this kind are quite large (composed of several dozen to several hundred layers), and large computational resources are required for training. Every picture, however, consists of approximately the same set of elementary objects: points, sloping lines, fine detail resolutions, and so forth. As a result, one need not begin training the neural network from scratch every time, working instead from the last point.

For example, you can take a database of X-ray images of lung cancer and retrain only the last few layers of an existing neural network. And more importantly, additional training does not require powerful computing resources.

Each day brings with it more examples of the successful application of artificial intelligence in medicine: neural networks can already recognize malignant tumours, including skin neoplasms, blood clots, and visual impairments, and they are able to diagnose the condition of internal organs on the basis of ultrasound, x-rays, MRIs, etc.

Weak artificial intelligence technology has been an extremely popular topic lately, though some aspects continue to fly under the radar. In the next issue, we will take look at these issues together in more detail. Stay tuned for more!

Material prepared by:
Vitaly Milke
Advisor to the President for Economy and Finance, JSC Business Alliance
PhD reseacher in Computer Science & Machine Learning

Articles on the topic
Analytics on the topic
Interesting on the portal
Regional economy between January and September 2019
The publication by the Institute for Complex Strategic Studies (ICSS) provides an analysis of economic indicators of Russian regions over the period from January to September 2019.
Prospects for the Use of Gas-Motor Fuel in Russia
The call of economic growth and economic development requires diversification of different types of fuel used in car industry, marine and railway transportations. By the volume of stocks, maturity of technology, influence on ecology, self-cost for final customers gas fuel has a lot of advantages in front of hydrogen fuel and electrical engines and gas fuels can be most demanded fuel for the nearest decade. Low cost can be the basis for economic growth compensating growing cost of petrol and diesel fuel. Additional positive effect is safety for ecology. Major problems the industry of gas fuel encounters are in poor infrastructure, insufficient legislation and high cost of gas engines. Future development of gas fuel industry is only possible with state-based subsidiaries with priority to development of infrastructure.
Expert opinion
Kirill Komarov: “Nuclear technology is much more than just electricity”
Kirill Komarov, Senior Deputy General Director — Director, Development and International Business Unit, Rosatom State Corporation, on why many African countries are considering joining the nuclear club.
Analytical digest
Luxury of event industry: F2F interactions as investments