Crowds during our talk on applying deep learning techniques to retail analytics. Our solution combined region-based convolutional Neural Networks and NLP techniques. Photo courtesy of deepsense.ai

Warsaw, Poland and Palo Alto, CA, USA - March 21, 2019

by Konrad Budek, deepsense.ai


deepsense.ai, Google Brain, the University of Warsaw and the University of Illinois at Urbana-Champaign have concluded a collaborative research project, building neural networks that mimic a simulated environment and effectively enabling artificial intelligence to perform a simulation.

“The lack of knowledge and courage to adopt Machine Learning models in daily business processes is the greatest obstacle companies face in leveraging this outstanding tool of growth.” Photo courtesy of deepsense.ai

Designed models can be applied in text or video prediction and to solve the problems where mathematical description may be overly complicated.

Photo courtesy of deepsense.ai

Reinforcement learning (RL) enables a neural network to develop and hone its skill by interacting with its environment, which is usually a simulated one.

https://deepsense.ai/what-is-reinforcement-learning-the-complete-guide/  

SpiNNaker - the first human brain supercomputer - it builds a working, small-scale model of the human brain, with the manifold scientific possibilities it brings with it. Read our AI monthly digest to stay updated! Photo courtesy of deepsense.ai

The technique is predicted to shine in robotics and the building of autonomous vehicles.

The research team found a way to build a neural network that can mimic the signals that the RL agent usually gets from interacting with the environment.

The neural network produces signals that usually would be taken from sensors, like images.

deepsense.ai‏ at AI World Expo in Boston. Let's talk what steps your company should take next to accelerate AI initiatives. Photo courtesy of deepsense.ai

“This is one of the important ideas of reinforcement learning. A recent survey done by DeepMind’s J.B. Hammrick provides a thorough account of analogies between model-based reinforcement learning and mental simulation as considered by cognitive science”, says Henryk Michalewski, the R&D Coordinator at deepsense.ai and a visiting professor in the Computer Science Department of the University of Oxford.

https://www.sciencedirect.com/science/article/pii/S2352154618301670  

Team deepsense.ai‏ at Vision in Stuttgart. Learn more about AI powered computer vision. Photo courtesy of deepsense.ai

 
The challenges of education

In the learning process, the RL agent is rewarded for performing a task correctly and punished for making mistakes.

Autonomous cars provide perhaps the best example–the agent is rewarded for safe driving and punished for collisions and speeding.

deeplearning for autonomous racing - learn by playing! Get details of the new training. Photo courtesy of deepsense.ai

The model seeks to maximize the rewards and minimize the punishments.

Initially, all actions are random and the neural networks have yet to explore the possible ways to perform the tasks.

Photo courtesy of deepsense.ai

It is thus possible that more than the first dozen rides will end with a collision with the nearest wall as the model incurs punishment and learns how to brake or avoid it.
This leads to the idea of conducting the trial and error process inside a simulator.

Photo courtesy of deepsense.ai

“However, building a realistic simulator is a tedious problem for humans” – says Błażej Osiński, a Senior Data Scientist at deepsense.ai.
“For example, Unreal Engine 4 has several million lines of code. The idea of our project is to let the neural network learn how to simulate the environment. A similar approach was suggested in a recent work from Yann LeCun’s lab, where neural networks were employed to simulate dense traffic.”

https://arxiv.org/abs/1901.02705  

Online course vs. instructor-led training compared. Photo courtesy of deepsense.ai

The research team has built a neural network that emulates the Atari gaming environment, a popular training ground for reinforcement learning models.

The network was able to create a version of the games Pong, Freeway and others that were nearly indistinguishable from Atari’s.

Photo courtesy of deepsense.ai

Ongoing research

The project was run by data scientists and researchers from deepsense.ai including Henryk Michalewski, Piotr Miłoś, Błażej Osiński and fellow researchers from Google Brain, the University of Warsaw and the University of Illinois at Urbana-Champaign.

Photo courtesy of deepsense.ai

More detailed information about the research, outcomes and possible uses can be found in the Arxiv paper and a detailed blogpost about artificial imagination on deepsense.ai’s blog.

https://arxiv.org/abs/1903.00374  

https://deepsense.ai/building-a-matrix-reinforcement-learning-and-artificial-imagination/  

Photo courtesy of deepsense.ai

The research opens up interesting ways to apply neural networks in business settings.

The first is to build an RL agent that can explore a highly complex environment and then emulate it for the needs of models that are to perform tasks within it.

The best example may be teaching the model to emulate the real world in all of its complexity and unpredictability.

Photo courtesy of deepsense.ai

Another business application would be a video prediction tool that could deliver videos based on only a few frames, effectively reducing the effort animators and designers put into producing full-length videos.

Photo courtesy of deepsense.ai

“We consider our research work an essential part of deepsense.ai and a key aspect of the company’s development. This time our researchers have really pushed the boundaries of knowledge with no marketing overstatements. The project brings fresh, new idea into AI research, a thing that is worth contributing into.”– concludes Tomasz Kułakowski, CEO at deepsense.ai.

Photo courtesy of deepsense.ai

About deepsense.ai:

deepsense.ai helps companies gain competitive advantage by providing them with AI-based end-to-end solutions, with the main focus on computer vision, predictive analytics and NLP.


We also deliver machine learning and deep learning training programs to support companies in building AI capabilities in-house.

https://deepsense.ai/  

We use methodology developed during research and commercial projects we have done for clients including NTT, Nielsen, L’Oreal, Google and Intel.

Photo courtesy of deepsense.ai

Media Contact:

Konrad Budek
konrad.budek@deepsense.ai  

Photo courtesy of deepsense.ai

Photo courtesy of deepsense.ai

Source: deepsense.ai

https://deepsense.ai/  



ASTROMAN Magazine - 2018.11.15

Fourth industrial revolution and machine learning

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2621  


ASTROMAN Magazine - 2018.09.01

How machine learning improves business efficiency - five practical examples

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2576  


ASTROMAN Magazine - 2018.09.01

Four ways to use a Kaggle competition to test artificial intelligence in business

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2572  


ASTROMAN Magazine - 2018.04.10

Financial Times: CodiLime among 50 fastest growing companies in Europe

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2480  


ASTROMAN Magazine - 2018.03.23

deepsense.ai becomes NVIDIA Deep Learning Partner

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2469  


ASTROMAN Magazine - 2018.03.14

Research on reinforcement learning advancements powered by deepsense.ai granted with 500,000 USD

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2458  


ASTROMAN Magazine - 2018.03.01

CodiLime's Training & Development Hub to help companies develop unique skills in-house

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2450  


ASTROMAN Magazine - 2018.01.05

United Nations enlists deepsense.ai to build out its deep learning expertise

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2406  


ASTROMAN Magazine - 2017.10.25

deepsense.ai popularize machine learning at European universities as part of the Intel Nervana AI Academy

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2367  


ASTROMAN Magazine - 2017.09.28

Seahorse goes open source! Data analysts can get more from the free BI tool powered by Apache Spark

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2345  


ASTROMAN Magazine - 2017.09.26

CodiLime's cybersecurity team among five best in the world

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2342  


ASTROMAN Magazine - 2017.09.14

deepsense.ai's research on robotics at prestigious 1st Annual Conference on Robot Learning

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2337  


ASTROMAN Magazine - 2017.09.06

The new Neptune. Data scientists on a cloud with the new Machine Learning Lab from deepsense.ai

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2331  


ASTROMAN Magazine - 2017.08.25

deepsense.ai has become the AI World's Strategic Deep Learning Partner

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2321  


ASTROMAN Magazine - 2017.08.19

deepsense.ai: A new global player on the AI solution provider market

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2316  


ASTROMAN Magazine - 2017.06.06

CodiLime's CodiSec CTF team beats out more than 200 competitors from across the globe in Göteborg, Sweden

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2270  


ASTROMAN Magazine - 2017.05.23

Intel brings data science to Polish universities with an event series powered by deepsense.io

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2254  


ASTROMAN Magazine - 2017.05.19

GPU Technology Conference 2017 - deepsense.io shares first-look on image recognition for retail analytics using deep learning

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2249  


ASTROMAN Magazine - 2017.05.05

CyCon 2017 Starts in Four Weeks. Czech Team Wins Cyber Defence Exercise Locked Shields 2017

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2239  


ASTROMAN Magazine - 2017.03.24

deepsense.io helps improve the UK's satellite defence and security intelligence

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2218  


ASTROMAN Magazine - 2016.10.21

CodiLime is second fastest-growing tech firm in Central Europe - 2016 Deloitte Central Europe Technology Fast 50 list announced

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2140  


ASTROMAN Magazine - 2016.10.20

The Deloitte Technology Fast 50 in Central Europe: First place in the 2016 ranking goes to the Polish business Codewise, the second-placed CodiLime also from Poland

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2139  


ASTROMAN Magazine - 2016.05.17

Tomasz Kulakowski, CodiLime co-founder and CEO, wins Poland's top business award for Vision and Innovation

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2066  


ASTROMAN Magazine - 2016.04.07

deepsense.io experts to present Big Data and deep learning accomplishments at Silicon Valley and Dublin conferences

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=2051  


ASTROMAN Magazine - 2015.10.31

CodiLime has won the Rising Stars Award and was ranked 2nd in Central Europe at the 2015 Deloitte Technology Fast 50 in Central Europe

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=1987  


ASTROMAN Magazine - 2015.10.23

deepsense.io Data Scientist Wins AAIA'15 Data Mining Competition

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=1983  


ASTROMAN Magazine - 2015.09.30

deepsense.io to feature Seahorse Community Edition on the Trusted Analytics Platform developed by Intel

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=1971  


ASTROMAN Magazine - 2014.12.12

CodiLime CEO, Tomasz Kulakowski became the EY Entrepreneur Of The Year 2014 in the category New Business

https://www.astroman.com.pl/index.php?mod=magazine&a=read&id=1843