Learning binary image representations and the path towards 3D

Learning a compact and efficient representation of images, words or 3D shapes is critical to many real-life applications, including visual search, 3D reconstruction and object recognition. In this talk, we will present several methods that address this problem, starting from binary image descriptors learned with linear projections to unsupervised training of 3D point cloud representations built with Generative Adversarial Networks to be applied in the context of autonomous cars. We will start wi...

Tomasz Trzciński is an Assistant Professor at Warsaw University of Technology and Chief Scientist and Partner at Tooploox. He received his PhD in computer vision from EPFL, previously worked at Google and Qualcomm and held a Visiting Scholar position at Stanford University. Member of IEEE and Computer Vision Foundation, he frequently serves as a committee member fo major vision and machine learning converences (CVPR, ICCV, ECCV, NeurIPS). Recently co-founded Comixify, VC-backed startup that uses GANs and reinforcement learning to turn your video into comics.

Sound event detection using self-consistency checking neural networks

Sound event detection is a challenging task even for most sophisticated neural networks. Effective merging of local and global context of time series is the key for achieving good performance. In my talk I will present an interesting new approach for utilization of global context, developed by my team at StethoMe. The method introduces new definition of training target for audio events that is inspired by Capsule Networks proposed by Geoffrey Hinton and residual training. Examples will include r...

Tomasz Grzywalski: Researcher, developer and tester of cutting-edge machine learning algorithms for sound & image processing. Currently works as a Lead Machine Learning Specialist at StethoMe where he, together with his team, develops algorithms for for automatic detection and classification of lung and heart sounds. In 2018 his team won 2nd place in FEMH ​Voice Data Challenge, a competition for diagnosing voice disorders based on vowel sounds.

Scale-invariant online learning

The talk concerns online learning with linear models, where the algorithm predicts on sequentially revealed instances (feature vectors), and is compared against the best linear function (comparator) in hindsight. Popular algorithms in this framework, such as Online Gradient Descent (OGD), have parameters (learning rates), which ideally should be tuned based on the scales of the features and the optimal comparator, but these quantities only become available at the end of the learning process. We...

Wojciech Kotłowski, PhD is an Assistant Professor at Poznań University of Technology. From 2009 to 2012, he was a post-doctoral researcher in Centrum Wiskunde & Informatica (Amsterdam, Netherlands) in the group of Peter Grünwald. His main research interests are in the theory of machine learning, particularly in the online learning with adversarial data. He is the author of many publications on top-tier machine learning conferences such as NeurIPS, ICML and COLT.

How to find that red dress? - Visual Search in Allegro

Jak skutecznie znaleźć na Allegro koszulę widzianą ostatnio w sklepie? Albo te nietypowe filiżanki, w których podają kawę w kawiarni na dole? Można oczywiście próbować opisać je słownie, ale czy nie byłoby prościej gdyby wystarczyło zrobić im zdjęcie? Podczas tej prezentacji przedstawię różne modele używane do wyszukiwania obrazem, opiszę jak one działają oraz pokażę przykładowe wyniki. Będzie dużo obrazków :)

Piotr Rybak pracuje jako Senior Research Engineer w zespole Machine Learning Research w Allegro i zajmuje się rozwojem modeli do wyszukiwania obrazowego. Absolwent matematyki, pasjonat uczenia maszynowego, miłośnik gier planszowych i wspinaczki.

The machine learning for de novo drug design

The Recurrent Neural Networks (RNNs) are capable of solving the problems involving some sort of sequence. Actually, the sequence can be considered as the context prototype. This means that not only the observation recorded at certain point carries the content, but also the relation with the neighbors brings significant piece of knowledge. One can mention a substantial number of use cases where the same computational techniques, i.e. RNNs, are employed in order to solve semantically very distant...

Rafał A. Bachorz, PhD is Data Scientist at PSI Poland and Head of Molecular Modeling Laboratory, Institute of Medical Biology, Polish Academy of Sciences.

Is practical AutoML more than CASH?

For a broader adoption and scalability of machine learning systems, the construction and configuration of machine learning workflow need to gain in automation. In the last few years, several techniques have been developed in this direction, known as AutoML. AutoML focuses on solving the black-box optimization problem named CASH, for Combined Algorithm Selection and Hyperparameter optimization. As its name states, CASH consists in selection the proper algorithm and its set of hyperparameters. H...

Alexandre Quemy joined IBM in 2015 where he currently holds a senior engineer position in the Data and AI organization. Prior joining IBM, he worked at Inria research center in France, in the team Machine Learning and Optimization on multi-objective optimization and meta-heuristics. In parallel of his activities at IBM, he is pursuing a Ph.D. at PUT, under the supervision of Pr. Wrembel, in the field of AutoML in unstructured space with a focal point on explainability.

Scaling machine learning systems up to billions of predictions per day

Whether it's a linear regressor or a system of connected deep learning models, getting your models ready is half the battle. Did you design your machine learning system to survive the onslaught of visitors from your latest Reddit and Hacker News post? Or the influx of users shopping during Black Friday? Are you ready for a world filled with flakey networks, invalid data, and impatient users? In this talk you'll learn how to design and architect your machine learning systems for the harsh realiti...

Carmine Paolino is a Senior Data Scientist at OLX Group where he works on computer vision, interpretable machine learning, and recommender systems. He has more than 4 years of industry experience as a Data Scientist, and more than 9 as a Software Engineer. Soon after his bachelor thesis he published a book chapter about large-scale distributed graph analysis, and he has recently collaborated on a state-of-the-art saliency detection technique. He's also an avid open source contributor, with patches to MXNet Model Server and Theano amongst others. When he's not working, he loves to take photos, write music, DJ, and learn more about Machine Learning, psychology, and the stock market. Once saved the Coding Horror blog from losing all its content.