As a graduate student in neuroscience at the University of California at Berkeley, Bilenko met students Asako Miyakawa and James Gao, who would become co-creators of Dr. Brainlove and the Cognitive Technologies exhibit. She became a researcher in Professor Jack Gallant's lab, collaborating with him on a project to train a computer to reconstruct a person's visual experience by interpreting EEG signals from the brain. The researchers created a computer algorithm that could interpret the brain impulses, match them against visuals collected from thousands of hours of YouTube videos, and generate movies that illustrate likely images based on the impulses. To outside observers, the results looked freakishly like computer-assisted mind reading. While the resulting visualizations were hazy, the work made a big splash in the tech community. It even prompted Mark Zuckerberg to declare that in the future "you're going to be able to capture a thought."

Many in Russia, in fact, seem to prefer their AI helpers with sass and a dark sense of humor. The Moscow-based tech giant Yandex launched a Russian-speaking personal assistant called Alice this October (pictured above, sans snark). And unlike Siri or Alexa, the program relies less on scripted responses than on what it's learned by consuming conversational data mined from the Web, news articles, and even a little Russian literature. As a result, Alice can respond to a much wider range of queries. However, some of program's responses can be a bit surprising.

Machine learning helps make decisions by analyzing data and can be used in many different areas, including music choice and facial recognition. Yandex, one of Russia's leading tech companies, has made its advanced machine learning algorithm, CatBoost, available free of charge for developers around the globe. "This is the first Russian machine learning technology that's an open source," said Mikhail Bilenko, Yandex's head of machine intelligence and research. What do cats have to do with this?

In many applications such as recommender systems and multi-label learning the task is to complete a partially observed binary matrix. Such PU learning (positive-unlabeled) problems can be solved by one-class matrix factorization (MF). In practice side information such as user or item features in recommender systems are often available besides the observed positive user-item connections. In this work we consider a generalization of one-class MF so that two types of side information are incorporated and a general convex loss function can be used. The resulting optimization problem is very challenging, but we derive an efficient and effective alternating minimization procedure. Experiments on large-scale multi-label learning and one-class recommender systems demonstrate the effectiveness of our proposed approach.

In this paper, we present a constrained co-clustering approach for clustering textual documents. Our approach combines the benefits of information-theoretic co-clustering and constrained clustering. We use a two-sided hidden Markov random field (HMRF) to model both the document and word constraints. We also develop an alternating expectation maximization (EM) algorithm to optimize the constrained co-clustering model. We have conducted two sets of experiments on a benchmark data set: (1) using human-provided category labels to derive document and word constraints for semi-supervised document clustering, and (2) using automatically extracted named entities to derive document constraints for unsupervised document clustering. Compared to several representative constrained clustering and co-clustering approaches, our approach is shown to be more effective for high-dimensional, sparse text data.

Semi-supervised learning is an active research field. Previous results shown that unite background information into the original unsupervised clustering problem could archive higher accuracy. In this paper, we explore the cooperation between the pairwise constrains given by the user and the sematic information in natural language. In addition, we reduce the time complexity to make the algorithm feasible for large quantities of data. Experiments on different scales of corpus show the robustness and effectiveness of the proposed algorithm, which the F-measure archives 20% higher than previous algorithms.