Industrial lectures

Vitaly Shiryaev, NLMK, “Data analysis and mathematical modeling on NLMK”

Abstract. The lecture will briefly outline the scope of production problems successfully solved at Novolipetsk Steel that are associated with the implementation of machine learning, mathematical modelling and optimization algorithms. The primary focus will be on the two following problems – Optimization of the Thermal Power Station (TPS) Performance and the Prediction of the Work Rolls Wear in the Hot Strip Mill.
The first problem is associated with the reduction of the natural gas procurement that is used for the electricity generation and steam production. This problem is addressed by achieving the optimal load on the TPS boilers and the maximum utilization of the coke and blast furnace gases, which are the byproducts of the metallurgical processes at NLMK. The optimal load takes into account that the boilers efficiency depends on the utilized gases proportion, as well as on their pressure and other factors. The dependence of vapor production on natural gas consumption by each boiler is found from the historical data by using advanced statistical methods.
As for the second problem focused on the work roll wear in the hot strip mill, the participants of the Summer School are invited to solve it during the extramural two-day competition. The competition results will be announced on the last day of the Summer School.

Irina Piontkovskay, Huawei, “Projects and Research Directions of Huawei Noah’s Ark Lab”

Tutorials and workshops

Mikhail Korobkin, Yandex, “HD maps construction”

To participate in this workshop, please, register here.

Abstract. We give an introduction to probabilistic robotics. Take a tour to GraphSLAM optimization and building High Definition (HD) maps. We are going to learn how to construct basic probabilistic control and perception models aligned with raw sensor data and then find parameters of the joint model via optimization.

Anton Dvorkovich, Yandex, “Correcting typos”

Konstantin Yakovlev, AIRI of FRC CSC RAS, “Single- and multi-agent path finding algorithms”

Prerequisites: Python+Jupyter. A*/CBS Jupyter notebooks are here.
Abstract. Path finding is a vital capability of an intelligent agent (robot, computer game character, etc.) operating in either virtual or real environment. It has been extensively studied and although a profound progress was made in this field, solving numerous path planning problems is still a challenge.
It this tutorial we will cover two the most prominent path finding algorithms, i.e. A* for single agent PF and Conflict Based Search (CBS) for multi-agent PF. We will go through the main theoretical concepts that lie behind those algorithms and will code them in Python.

Ricardo R. Gudwin, “Using CST to build a Cognitive Architecture controlling an NPC in a Computer Game”

Software and task description are here

Abstract. We provide a step-by-step demonstration illustrating the main foundations of the CST Cognitive Systems Toolkit in building a cognitive architecture to work as an artificial mind for controlling an NPC (non-player character) in a 3D virtual environment computer game. We start by introducing the main foundations of CST: Codelets and Memories, and how they should be used to integrate a cognitive architecture, controlling the NPC. The demonstration is a hands-on programming activity, using Java and Netbeans as language/tool.

Dilyara Baymurzina, iPavlov, “How to solve NLP tasks with DeepPavlov” 

To participate in this workshop, please, register here.

Prerequisites. Python+Jupyter.

Abstract. DeepPavlov is the open-source library for building chat-bots and complex conversational systems. The introductory lecture on existing NLP tasks and current common solutions will be given. The second part of the workshops is to familiarize participants with DeepPavlov main commands and to try to use several DeepPavlov components.

Ivan Fursov, Tinkoff“Neural Networks with Attention Mechanism for Efficient Paraphrase Retrieval” 

To participate in the tutorial, plese, register here

Abstract. In this tutorial, we study the task of modeling a system, which aims to compare two sentences and identify the relationship between them. In particular, we will develop a paraphrase identification model, and use it to determine whether two sentences are paraphrase or not. The tutorial covers baseline approaches to solving this task, attention mechanisms in NLP, techniques to speed up inference and how a paraphrase identification task can replace text classification task.

Ivan Mazurenko, Huawei, “On fundamental mathematical problems of deep learning”

Abstract. A survey of open mathematical problems in the theory of deep neural networks (such as expressiveness, robustness, convergence, capacity problems) will be preented. Practical applications of these problems will be illustrated using examples from Huawei industry projects.

Ilya Shepel, NKB VS, “ROS and virtual modeling tools. How to start assembling and developing robots “on the table”.

To participate in the tutorial, plese, register here

Abstract. In this tutorial we will show how to develop the algorithms of a mobile robot with technical vision based on lidar. We will solve the problem of autonomous movement in the external environment using simulation tools and ROS infrastructure.


NLMK, “The prediction of rolls wear in the hot strip mill”

Full description on Kaggle

Prerequisites: Python, Anaconda (sklearn), Excel
Abstract. The metal processing in the hot strip mill is divided into the series of so-called rolling batches. After finishing each batch, the mill stops and the worn work rolls are replaced with the new ones. The removed rolls are then polished on the grinding machine in order to remove the surface defects and restore their original geometrical profile. After that, the rolls are installed back into the hot strip mill during its next stop.
The work rolls are utilized until their work surface, which is typically several centimeters thick, fully wears out. The wear is defined as the reduction of the roll diameter during its operation and the subsequent polishing. The wear depends on multiple factors such as the roll’s material, the roll’s position in the mill and the processed steel grade. The task is to create the mathematical model of the work roll wear by using the provided statistical data, as well as some physical reasoning. The model should be then used to predict the roll wear after processing each rolling batch.

CSD Lab, MIPT, “Computer vision and reinforcement learning”

Registration, Presentation, GitHub

Abstract. Creation of an improved program for the accident-free movement of a car at an unregulated crossroad using video from an unmanned aerial vehicle. This will require the usage of deep learning approaches to recognize cars on the images and reinforcement training to plan their movement.

Huawei, “Metric learning for facial descriptors”

Registration, Presentation, Kaggle, GitHub, Telegram

The problem: Construct and implement the metric (similarity function) which brings us the better percent of recognition (identification\verification) of human faces on photos in comparison to L2 baseline.

Prerequisites: Python 3.

Abstract: State-of-the-art recognition (identification) methods generally use the following paradigm: firstly vector (descriptor) of some object (e.g. photo) is calculated using deep neural net, and then using some simple metric (e.g. L2 in Euclidian space) we select the closest one(-s) descriptor from the preliminary computed set of descriptors of objects using the same neural net, among which we’ll search the appropriate pair request<->DB. For verification systems the metric is applied between the descriptor of object-request and the preliminary computed for this object descriptor-template, and a result conclusion on whether this is the same object or not is done based on comparison of this metric with some predefined threshold.