MLMU BA: Real (Scary) ML deployment stories – Vlado Boza
- Language: EN/SK
- Date and time: Wed, December 11th, 19:00
- Venue: ProgressBar, Dunajska 14
batching tricks and GPU magic ( on example where each word consists of the different number of subword units and you want to pool those representations).
Vlado might also discuss weird internal workings of matrix multiplication, convolutions and RNNs.Speaker:Vlado Boza has a PhD in bioinformatics from Comenius University. He is very passionate about efficient algorithms and optimization problems. Works at CEAI as ML lead. Currently, he is trying to detect cough from sound recordings. Won some money from Kaggle and is Competition Grandmaster there. Loves Rust. Dislikes Scala.
Facebook event: https://www.facebook.com/events/787616991680127/
Eventbrite ticket: https://www.eventbrite.com/e/mlmu-ba-real-scary-ml-deployment-stories-vlado-boza-tickets-84204414671
Meetup.com event: https://www.meetup.com/Machine-Learning-Bratislava-Meetups/events/266814080/
Event: Medical Image Retrieval – Rene Donner
Image retrieval systems in the medical domain support doctors in their decision making, by taking a query image or a region within such a query image as input and finding the most similar images/regions in vast databases. We will be looking at state of the art approches to build such systems, using deep learning for feature extraction and indexing.
With a background in electrical engineering René has worked for 8 years at the Medical University Vienna as a researcher in computer vision, focussing on anatomical structure localization and content based image retrieval. He is now CTO at contextflow, applying deep learning to large scale medical image data and developing smart tools to aid radiologists in their challenging tasks.
Eventbrite ticket: https://www.eventbrite.com/e/mlmu-ba-medical-image-retrieval-rene-donner-tickets-79970246159
Meetup.com event: https://www.meetup.com/Machine-Learning-Bratislava-Meetups/events/266187860/
Facebook event: https://www.facebook.com/events/633931594105430/
Event: How to do ML if you have lots of Google’s GPUs
Journey of a student from UK Bratislava into doing research at Google New York as an AI resident. What is AI residency, what is AutoML, what is architecture search, how can one make use of tons of GPUs, why does AdaNet have such an amazing GIF, what does it mean to play at state-of-the-art levels of accuracy in image classification, and what is so far the best indicator for getting into the AI residency?
Vladimir „Vlejd“ Macko is a graduated from UK Bratislava with 4 years of ML startup experience and two internships in Google. He spent his last year in Google research New York as an AI resident. He worked on architecture search in AutoML team AdaNet, and combinatorial optimization for mixed integer program solving in collaboration with DeepMind.
Facebook event: https://www.facebook.com/events/497961537723741/
Eventbrite ticket: https://www.eventbrite.com/e/mlmu-ba-how-to-do-ml-if-you-have-lots-of-googles-gpus-vladimir-macko-tickets-75600253393
Event: Real Estate Price Prediction
The talk will present our efforts over the past year and a half to launch a start-up for US real estate valuation. The start-up called Babcock & Bonbright was built as part of the Central Europe AI portfolio – a start-up studio focused on using artificial intelligence to innovate in the financial and healthcare sectors.
The presentation will focus mainly on the technological aspect of the project – linking structured data, NLP, and image processing to a comprehensive system for estimating the value of dwellings and apartments using machine learning. Attention will also be paid to the processing of map data, system architecture and data pipeline design.
Lukáš Vrábel finished his Master’s in Artificial Intelligence at the Brno University of Technology. There he continued to perform research and education in the area of theoretical computer science. Later he returned to machine learning at Seznam.cz – a local Czech competitor to Google – where he worked at multiple positions ranging from Data Scientist to Head of Research Department. He entered the startup scene by establishing and growing the Brno office for US-EU startup studio Central Europe AI. During his stay there, he also worked on machine-learning based real estate price prediction. Lukáš is currently working as a freelancer, helping various companies with building data science departments, teams, and projects.
Facebook event: https://www.facebook.com/events/597179320792819/?notif_t=event_calendar_create¬if_id=1559587728237461
Eventbrite ticket: https://www.eventbrite.com/e/mlmu-ba-real-estate-price-prediction-lukas-vrabel-tickets-62867381029
Meetup.com event: https://www.meetup.com/Machine-Learning-Bratislava-Meetups/events/262011451/?isFirstPublish=true
Pitfalls of Emotion detection in production
Implementing a deep neural network in production environments presents several challenges, including defining the architecture, training the models, and developing an application. This talk will demonstrate some of the challenges faced, and how you can navigate the terrain of deep learning in real world applications.
Statistical vs. Deep Learning Methods for Time Series Forecasting
In recent years, Deep Learning (DL) has revolutionized many fields such as image analysis, speech recognition, and natural language processing. However, Time Series Analysis is still dominated by classical statistical methods. In a recent comparison of statistical (Stat) and machine learning (ML) forecasting methods, prof. Makridakis, one of the leading authorities in this field, even claimed: „The forecasting accuracy of the best ML method was lower than the worst of Stat ones while half the ML methods were less accurate than a random walk“ (Makridakis et al., PLOS One, 2018).In this talk, we start by a high-level introduction to time series forecasting. Next, we get an overview of M1 – M4 competitions‘ results and publicly available datasets on Kaggle. We propose an explanation why for some data, DL forecasting methods are superior, while on the other datasets, they cannot compete with Stat methods. And what can help you to choose in an era of automatically generated time series all around.
Travelling salesman problem is NP-hard? No Problem!
But in this talk we take optimization seriously. I will talk about our experience in two Kaggle competitions (https://www.kaggle.com/c/traveling-santa-problem, https://www.kaggle.com/c/traveling-santa-2018-prime-paths), which were related to Travelling salesman problem (TSP). I will talk about main approaches for solving TSP and also how to show that solution is almost optimal. Then we discuss implications of variations encountered at above mentioned Kaggle competitions.
Vlado Boza has a PhD in bioinformatics from Comenius University. He is very passionate about efficient algorithms and optimization problems. Works at CEAI as ML lead. Currently, he is trying to detect cough from sound recordings. Won some money from Kaggle and is Competition Grandmaster there. Loves Rust. Dislikes Scala.
Anomaly detection using time series forecasting models in business
In the beginning we will talk about the theoretical part about time series, forecasting models and evaluations methods. After that, Jan will speak about the process of creating evaluation methodology for the case he was dealing with in Dell. Then he will mention how anomaly detection can be used for marketing purposes in examples such as media monitoring cases, that he was working on in Mediworx.
Jan Deak has a Statistical degree from University of Economics in Bratislava and works as a Data Scientist in Mediworx. He has more than 4 years of experiences working as Data Scientist (GFK, Dell, Mediworx). Mostly focused on text mining, AI, anomaly detection, CRM data, forecasting and classification problems.
How AI helps robots to see and manipulate objects (Michal Malý)
Humans are perfect at object handling. For a very long time, industrial robots were doing only simple, hard-coded manipulation. High-quality 3D scanning enabled us to localize objects precisely, and plan their picking and movement to avoid collisions. Localization and picking are tasks that we are currently solving with both analytical and ML approach. We compare the advantages and disadvantages of the approaches.
Michal (34) received his PhD in the field of Computer Science, Artificial Intelligence at FMFI UK in 2013. He is an expert in construction of rational agents based on reinforcement learning. He also worked in the area of computer security and published an innovative solution of secure distributed computing. He is currently Director of AI in Photoneo. His vision is to improve computer understanding of 3D environment and enhancing comprehension of real-world objects.
Towards Interfaces in Natural Language (Zuzana Nevěřilová)
Smart search as an essential part of the Konica Minolta Cognitive Hub needs to “understand” search queries in English. Being very different from what we use for searching the Web, smart search uses natural language processing and machine learning methods to detect the user’s intent. Afterwards, it extracts all the information the user mentions and creates relationships with the internal knowledge base. The talk will present methods Konica Minolta uses in the natural language query parser and discuss the challenges in creating natural language interfaces.
Zuzana Nevěřilová finished her PhD at Masaryk University in Brno in 2014. She works as a researcher at the Natural Language Processing Centre at Masaryk University where she focuses on knowledge bases and semantics. From June 2018, she is member of the Konica Minolta Laboratories Europe.
AI/ML for Scalable Recommender Systems (Jakub Mačina)
Recommender systems are successfully used in several domains such as e-commerce or multimedia recommendation. In the e-commerce field, increasing number of companies utilize recommender systems to offer a personalized shopping experience for their customers.
In this talk, we will provide an overview of typical recommender system architecture, approaches used in the real-world applications and we highlight the challenges encountered by industry practitioners. We will discuss basic approaches, such as matrix factorization and text mining, and move on to deep learning techniques for recommendation. In the real-world case study we will show how to capture attributes of products into a dense (latent) embedding representation. Finally we will focus on evaluation metrics and interactive learning for optimizing long-term revenue of retailers. Everybody with an interest in machine learning is welcomed.
Jakub Mačina is the Recommendations Team Lead at Exponea. He graduated from the Slovak University of Technology with specialization in machine learning and information retrieval. His research background is in the field of recommender systems with a publication at the premier conference on recommender systems research ACM RecSys. He is passionate about movement and open source software.
Exponea Experience Cloud is an award-winning customer experience and data management platform that boosts e-commerce growth with AI-powered engagement automation and helps improve company culture with improved cross-department collaboration and customer centricity.
Machine Learning Challenges in DNA Sequencing (Tomáš Vinař)
MinION is a pocket-sized device that plugs into a USB port of your computer and can decipher your DNA. The device measures changes in electric current as DNA molecules pass through nanopores, resulting in sequences of current measurements called squiggles. It turns out, that interpreting these squiggles is neither easy nor fast. In fact, machine learning methods are the key to bringing this device to Africa to monitor most recent disease outbreaks, or to the desk of your medical practitioner to determine personalized medical treatments.
Tomáš Vinař has studied PhD at the University of Waterloo in Canada, after which he has spent some time at Cornell University in USA as a postdoc. Currently, he is an associate professor and one of the principal investigators of the Computational Biology Research Group at the Faculty of Mathematics, Physics and Informatics of Comenius University in Bratislava. Besides doing bioinformatics research in collaboration with biologists, he teaches bioinformatics, machine learning, and algorithms.
Can deep neural networks be used on embedded devices? (Marián Beszédeš)
Usage of deep neural networks is often connected to powerful (multi)GPU machines. However, we want to use their potential on every piece of hardware, even on mobile phones. It is still not straightforward to use them on embedded devices because of their size, resource consumption, performance and the like.
Nevertheless, there are techniques which can be used to make deep neural networks and embedded devices more compatible. We will talk about specialized neural networks architectures, quantization, mobile platform optimized NN libraries and even more.
Using object detection with CNTK to classify hotel images (Karol Żak)
In our next talk you can look forward to a presentation that will reveal ML application in software products – our speaker Karol Żak will show you how they used object detection with FasterRCNN using CNTK to classify hotel images. All the learnings they share, come from cooperation with customer on real-life solution – platform Hotailor with nearly a million hotels inventory using smart analytics to help businesses sell more hotel rooms.
Karol Żak is Technical Evangelist @Microsoft – a technology enthusiast for years associated with Microsoft and IT communities in Poland. He is also former Microsoft Student Partner, founder and ex-leader of .NET Group in Technical University of Radom. Karol is all about new technologies and trends in IT but mostly interested in cloud computing, AI/ML and mobile tech.
For 2 years as a certified trainer with MCT title, he led IT trainings for developers and then in 2014 joined Microsoft as a Technical Evangelist. Privately he is a big american football fan and starting quarterback for one of Warsaws teams and also Polish National Team.
Implications of adversarial environment on machine learning (Michal Nánási)
Speech data mining: Not Yet Ready for Retirement (Honza Černocký)
Time series prediction with Neural networks: Applications, Pitfalls and Beyond (Rudradeb Mitra)
ML Engineer’s look at insurance (Peter Zvirinský)
Cesta vidiaca autá, ktoré po nej jazdia (Adam Herout)
Okrem akadémie a biznisu Adam koučuje, a to hlavne profesionálov z oblasti IT.
3D Reconstruction in Wild (Martin Bujnak)
While it might look magical, there are still many opened questions and opened problems to solve.
He received his Ph.D. in computer vision from Center for Machine Perception in 2012 and the RNDr. and M.S. degrees in computer graphics and parallel computation from Comenius university in 2005. He pioneered and published several minimal solvers for camera pose estimation on major computer vision conferences and journals. He gained development skills in Caligary, Micrsoft live labs and Xbox.
Machine learning in adversarial environments for malware detection (Jakub Debski)
Seznam.cz – Vyhľadávanie obrázkov z (nielen) českého internetu (Lukáš Vrábel)
Využitie machine learningu v tvárovej detekcii a verifikácii (Marián Beszédeš)
The role of optimization in Machine Learning (Dominik Csiba)
Optimization is a key component of most of the machine learning methods. Current problems often feature large amounts of high-dimensional data fitted by a complex non-convex model, such as deep learning, which are very difficult to learn efficiently. Novel approaches arise every month to tackle these problems, but the training time is often still unsatisfactory. In this talk, I will present a few scenarios, where optimization is the main bottleneck – featuring questions such as: “What kind of problems are difficult for deep learning?”. The talk is aimed for general technical audiences, everybody is welcome!
Automatické budovanie modelov pre časové rady s aplikáciou v energetike (Ján Dolinský)
Ján Dolinský zo spoločnosti tangent.works vysvetlí, prečo je automatické budovanie modelov tak atraktívne pre priemysel a prečo sa v tangent.works rozhodli pretaviť vedomosti a skúsenosti do produktu, ktorý sa nazýva TIM – Tangent Information Modeller. Ján bude rozprávať aj o informačnej geometrii, ktorá umožnila vybudovať tento produkt.
Extrakcia diskriminačných kľúčových slov (Márius Šajgalík)
Márius Šajgalík momentálne robí PhD v odbore Inteligentné informačné systémy na FIIT STU v Bratislave. Jadrom prezentácie bude reprezentácia textových dokumentov vo forme kľúčových slov pre účely kategorizácie a nové architektúry neurónových sietí, ktoré možno využiť nielen na extrakciu kľúčových slov bez učiteľa.
Budovanie NLP klasifikátorov (Vlado Boža)
Vlado Boža momentálne robí PhD v bioinformatike, účí ML na FMFI a okrem toho pôsobí ako machine learning lead v Black Swan Rational a aktuálne aj v CEAI Slovakia. Príde porozprávať o budovaní NLP klasifikátorov – od logistic regression po deep networks, od riešenia Kaggle súťaží po riešenei aktuálnych problémov v CEAI. Hovoriť bude aj o distant supervision a získavaní trénovaích dát a ukáže nám niekoľko efektívnych trikov na uľahčenie roboty.
Máš nápad na prednášku? Ozvi sa nám na email@example.com