Big Data Science

2021-07-16 05:24:58 AI will schedule flight crews for the US Air Force
The AI system from MIT helps US Air Force pilots plan the workload of personnel on cargo flights, based on many factors: airspace availability, pilot tolerances, work and rest requirements, etc. Combining optimization through integer programming with RL neural networks, the system generates flight schedules based on explicit and implicit constraints. https://news.mit.edu/2021/us-air-force-pilots-artificial-intelligence-assist-scheduling-aircrews-0708 Open / Comment

2021-07-14 05:58:47 Why is it raining not as predicted and how Yandex Meteum 2.0 deals with it
The story about replacing MatrixNet with CatBoost and new datasets for training NN models. Now Meteum neural nets learn not only on data from professional weather stations, but also on information about terrain features and user messages. https://tekdeeps.com/yandex-has-launched-meteum-2-0-a-new-technology-for-weather-forecasting-based-on-machine-learning/ Open / Comment

2021-07-12 05:17:21 ML for prediction of Solar Radiation
From a practical agronomic point of view, an accurate assessment of solar radiation is vital because it is a key factor in crop development. Most existing weather stations around the world have temperature and rain sensors, but only some of them measure solar radiation. Measuring solar radiation is usually very expensive due to complex sensors (pyranometers and radiometers) and a lack of reliable data. Therefore, a group of researchers from the University of Cordoba has developed ML-models to predict solar radiation in southern Spain and the United States.
The created ML-models are based not only on actual measurements, but are enriched with data on the geoclimatic conditions of the area (aridity, distance to the sea, altitude, etc.). To estimate daily solar radiation, the proposed neural network algorithms from current data only need information about the air temperature, which is relatively cheap due to inexpensive sensors and IoT technologies. Bayesian algorithms are used to optimize hyperparameters, and the models themselves can be adapted to any terrain, depending on its aridity.
https://techxplore.com/news/2021-07-machine-based-thermal-solar.html Open / Comment

2021-07-09 08:49:24 Need a quick rewrite? Try NLPAug!
NLPAug is a Python library that allows you to increase the efficiency of neural networks in NLP tasks without changing their architecture and fine-tuning. With it, you can synthesize new text based on the available data, replacing some words with synonyms, incl. by the principle of cosine similarity in vector representations, similar to word2vec or GloVe. NLPAug also performs context-based word replacement using transformers in the form of BERT networks and makes double translation of text into another language and vice versa. https://github.com/makcedward/nlpaug Open / Comment

2021-07-07 09:53:41 NetHack Challenge at NeurIPS 2021 from Facebook: open-source project for Reinforcement Learning (RL) as a game
Many advances in RL have been achieved through simulation environments in games such as Dota 2, Minecraft, and StarCraft II. But this requires a lot of computation on thousands of GPUs at a time for just one experiment. To reduce the cost of RL modeling, Facebook in 2020 initiated the development of the open-source NetHack Learning Environment project. And in 2021, Facebook announced the NeurIPS 2021 competition as part of the NetHack Challenge in conjunction with AIcrowd, an AI crowdsourcing organization. The competition runs from early June to October 15, 2021, and the winners will be announced on NeurIPS in December.
The NetHack game has actually existed since the 1980s. It is visually straightforward and completely free to play, but more complicated than StarCraft II due to the very confusing interaction of players with their environment and related objects, users have to think outside the box or refer to the NetHack Wiki. And the main difficulty of NetHack is that after the death of a character, the game session of this player ends. Therefore, within this RL environment, researchers hope to find new ways to control agents so that in the future, AI can think creatively in difficult situations, helping people. Because NetHack runs on a terminal, players can quickly simulate gameplay by training billions of agents a day on just 2 GPUs. This is how the NetHack Challenge tests the latest AI techniques in a complex environment without the enormous power of a supercomputer. https://techxplore.com/news/2021-06-facebook-nethack-neurips.html Open / Comment

2021-07-05 07:40:20 Game theory as an engine for large-scale data analysis
A new look at principal components analysis as a competitive game, where each approximate eigenvector is controlled by the player, whose goal is to maximize his own utility function. As a multi-agent perspective, it has allowed the development of new ideas and algorithms with the efficient use of the latest computing resources, globally scaling datasets. Brief overview https://deepmind.com/blog/article/EigenGame and detailed article https://openreview.net/forum?id=NzTU59SYbNq Open / Comment

2021-07-02 06:11:33 Preparing for code review, speeding up development and testing with Copilot for Visual Studio from OpenAI
Write faster and better in Python, JavaScript, TypeScript, Ruby, Go, and a dozen other languages. Copilot runs on Codex, a new AI system from OpenAI, and understands more context than IDE helpers. Whether it's a docstring, a comment, a function name, or the code itself, Copilot uses context and synthesizes the instructions you need to help the developer create a quality product. https://copilot.github.com/ Open / Comment

2021-06-30 18:01:08 10 the most interesting DS-conferences all over the world in July 2021
• 11.07 - MDA 2021: 16th International Conference on Mass Data Analysis of Images and Signals with Applications in Medicine, Biotech, and more. New York, NY, USA.
• 12.07 International Conference on Mobile Geomatics and Geodata Science (ICMGGS) - Ottawa, Canada https://waset.org/mobile-geomatics-and-geodata-science-conference-in-july-2021-in-ottawa
• 14.07 MLCon - The AI and ML Developer Virtual Conference, online https://cnvrg.io/mlcon/
• 18.07 - MLDM 2021: 17th International Conference on Machine Learning and Data Mining. New York, NY, USA http://www.mldm.de/
• 18.07 - International Conference on Machine Learning 2021. Online https://icml.cc/
• 20.07 - Chief Data & Analytics Officers (CDAO) Government Live, leading data-driven transformation across the government sector by Corinium. Online https://cdao-gov.coriniumintelligence.com
• 21.07 - Subsurface LIVE Summer 2021 - Free and Virtual https://www.dremio.com/subsurface/live/
• 22.07 - Business of Data Festival – Online https://www.boddigitalbroadcast.com/festival/home
• 22.07 - MarketsandMarkets Big Data Virtual Summit. Online https://events.marketsandmarkets.com/2nd-edition-marketsandmarkets-big-data-virtual-summit/
• 26.07 - The 2021 International Conference on Data Science (ICDATA'21). Las Vegas, NV, USA https://icdata.org/ Open / Comment

2021-06-30 06:28:33 The future of AI chips: a long history of collaboration and confrontation between AI giants to develop DS solutions. NVIDIA vs Google, the prospects for the development of deep neural networks and microcircuits, as well as money, cats and the Internet. https://www.wired.co.uk/article/nvidia-ai-chips Open / Comment

2021-06-28 05:45:37 One word is enough: Facebook AI announced a new project, TextStyleBrush, capable of copying the style of handwritten or typed text on an image using only one word. This is useful for changing and replacing text in photos. Unlike previous AI systems capable of copying text from photographs, TextStyleBrush can work with all types of calligraphy and typography, interpreting various rotations and transformations of text, from curved characters to the natural deformation of paper from pressing a pen. It is important that Facebook understands the possibility of misuse of the TextStyleBrush for malicious acts such as text deepfakes. To prevent such attacks, the company will share research results with the Deepfake Detection Challenge dataset, contributing to the broad knowledge base of DL-fakes.
https://techxplore.com/news/2021-06-facebook-ai-word-mimic-text.html Open / Comment