Paweł Gora is the founder and leader of an independent research group called TensorCell who are working on optimising complex processes using AI. They are currently focusing on developing AI algorithms for optimal traffic signal control and optimising cancer treatment using radiotherapy. Despite the fact that these processes are from different domains, they share many interesting properties and can be tackled using similar techniques. TensorCell's methods are mainly based on AI and multiagent simulations based on cellular automata, hence the name TensorCell (tensors are just mathematical objects which are extensively used in machine learning; cells are building blocks of cellular automata). They use cellular automata which can be highly parallelised to simulate complex processes, they train machine learning models to approximate the outcomes of the simulations in order to accelerate computations even faster, and finally, they use evolutionary algorithms and reinforcement learning to optimise the complex processes.

We apply advanced AI-based methods to optimise complex processes like road traffic or cancer treatment. Although these might seem totally unrelated, they share some similar processes and can surprisingly be addressed with similar techniques.

Paweł Gora

Founder & Leader, TensorCell Research Group

Computationally demanding datasets

TensorCell use data-driven methods to train advanced machine learning algorithms. To do so they need large datasets. Currently, such datasets can be prepared only using computationally-demanding simulations. Training machine learning algorithms and running experiments with optimisation algorithms are computationally-demanding tasks. Therefore, in the case of TensorCell, computational resources are vital in order to conduct experiments and research. But as an independent research group it is a challenge to find the necessary resources to continue their research.

OCRE’s CloudSigma cloud resources vouchers

Thanks to OCRE’s vouchers for CloudSigma’s cloud resources, the research group was able to generate large training sets for their new experiments. The flexibility provided by the vouchers helped TensorCell to prepare the datasets on time.

Our methods are data-driven - we train advanced machine learning algorithms for which we need large datasets. Thanks to OCRE’s vouchers for CloudSigma’s cloud resources, we were able to generate large training sets for our new experiments. Using cloud resources whenever we want gave us sufficient flexibility and opportunity to prepare the datasets on time.

Paweł Gora

Founder & Leader, TensorCell Research Group

New opportunities

Thanks to these generated datasets, TensorCell is able to conduct new experiments that could help solve existing traffic control and cancer treatment optimisation problems! Most of their datasets will be made publicly available in order to facilitate future research on these topics by other scientists.

For more information about TensorCell, click here.