Archive for the ‘Machine Learning’ Category

Reconstructing the Galactic merger history with machine learning – Astrobites

Title: Kraken reveals itself the merger history of the Milky Way reconstructed with the E-MOSAICS simulations

Authors: J. M. Diederik Kruijssen, Joel L. Pfeffer, Melanie Chevance, Ana Bonaca, Sebastian Trujillo-Gomez, Nate Bastian, Marta Reina-Campos, Robert A. Crain, and Meghan E. Hughes

First Authors Institution: Astronomisches Rechen-Institut, Zentrum fur Astronomie der Universitat Heidelberg

Status: Published in MNRAS [open access]

Just like archaeologists can trace the migration and assimilation of people in past societies, astronomers can reconstruct the assembly history of the Galaxy that we live in. In standard galaxy formation theory, galaxies like our Milky Way formed through the hierarchical merging of many smaller galaxies. According to this picture, some of the stars and star clusters in our Galaxy were not originally born here, but are immigrants that were brought into the Milky Way when their parent galaxy entered. Galactic archaeologists are developing techniques to trace back the origin of these galactic immigrants and reconstruct properties of the accreted galaxies. One avenue is through the stars that were left behind in a stream (see this Astrobite), but todays authors study where the star clusters in our galaxy come from.

Globular clusters consist of hundreds of thousands of tightly bound stars, and they are both ancient (billions of years old) and stable. When a satellite galaxy is accreted into the Milky Way, its globular clusters are likely to survive and migrate as a whole. This makes globular clusters excellent fossil records, because they preserve the metallicity of the environment in which they formed and carry this signature wherever they travel.

Astronomers can identify accreted globular clusters based on their age and metallicity. Figure 1, taken from todays paper, shows the observed age-metallicity relation for Milky Way globular clusters. The globular clusters on the main progenitor branch are clearly separated from those accreted from various satellite galaxies. The main progenitor of the Milky Way contains all the native stars and globular clusters that were not from satellite galaxies.

Figure 1. The age-metallicity distribution of Galactic globular clusters. In all panels, black points indicate globular clusters that formed in the Main progenitor while colored diamonds indicate globular clusters from each accreted satellite galaxy. The vertical line represents the inferred accretion time. Reproduced from Fig 3 in the paper.

The data points in Figure 1 only contain the observed properties of globular clusters, and there is no obvious connection to the satellite accretion events. To bridge this gap, the authors in todays paper make use of galaxy formation simulations. The E-MOSAICS simulations follow the co-formation and co-evolution of galaxies and their globular clusters, providing the crucial link between accretion history and globular cluster properties.

The authors train an artificial neural network to infer the progenitor galaxy that brought in a group of globular clusters. Specifically, the input parameters are the median and interquartile ranges (IQRs) of the globular cluster orbital radii, eccentricities, ages, and metallicities, and the networks can predict the accretion time and the galaxy stellar mass. The resulting neural network is applied to the globular clusters shown in Figure 1 and gives the accretion times as outputs.

After applying the artificial neural network, the reconstructed formation history of the Milky Way is shown in Figure 2. This figure is called a galaxy merger tree because each branch (black and grey lines) represents one accreted satellite galaxy, and the branches are ordered by their accretion times. Kraken was the first galaxy to be accreted, followed by the progenitor of the Helmi streams, Sequoia and Gaia-Enceladus, and finally Sagittarius.

Figure 2. Galaxy merger tree of the Milky Way. The main progenitor is denoted by the trunk of the tree, coloured by stellar mass. Black lines indicate the five identified (and likely most massive) satellites, with the shaded areas visualizing the probability distributions of the accretion times. The coloured circles indicate the stellar masses of the satellite galaxies at the time of accretion. The annotations list the minimum number of GCs brought in by each satellite. From left to right, the six images along the top of the figure indicate the identified progenitors, i.e. Sagittarius, Sequoia, Kraken, the Milky Ways Main progenitor, the progenitor of the Helmi streams, and Gaia-Enceladus. Reproduced from Fig 9 in the paper.

The main progenitor of the Milky Way is the trunk of the tree, and it grows in stellar mass each time it accretes a new galaxy. The thickness of the lines indicate the mass ratio of the accreted galaxy versus the main progenitor. As you can imagine, the more massive a satellite is, the more damage it causes when it combines with the Milky Way. In a minor merger (defined by mass ratios smaller than 1:4), the satellite is small enough for the Milky Way to comfortably absorb; however, in major mergers (where the two galaxies have comparable mass), both galaxies will be significantly disturbed and the Milky Way disk can even be destroyed. Luckily, the Milky Way never experienced a major merger according to the authors of todays paper. Among the minor mergers, Kraken was the most significant merging event that the Milky Way experienced, since it has the highest mass ratio at the time of accretion.

The authors tally up the total contribution of stellar mass and globular clusters from the accreted satellites. They find that only a few percent of the stellar mass and about 35-50% of globular clusters in the Milky Way were accreted. The rest formed inside the Milky Way. They conclude that the Milky Way had an unusually quiet formation history.

Todays paper uses globular clusters and artificial neural networks to reconstruct a detailed accretion history of our Galaxy. No Indiana Jones required for galactic archaeology!

Astrobite edited by Roan Haggar

Featured image credit: Diederik Kruijssen

About Zili ShenHi! I am a Ph.D. student in Astronomy at Yale University. My research focuses on ultra-diffuse galaxies and their globular cluster populations. Since I came to Yale, I have worked on two "dark-matter-free" galaxies NGC1052-DF2 and DF4. I have been coping with the pandemic and working from home by making sourdough bread and baking various cookies and cakes, reading books ranging from philosophy to virology, going on daily hikes or runs, and watching too many TV shows.

Read the rest here:
Reconstructing the Galactic merger history with machine learning - Astrobites

All The Machine Learning Libraries Open-Sourced By Facebook Ever – Analytics India Magazine

Today, corporations like Google, Facebook and Microsoft have been dominating tools and deep learning frameworks that AI researchers use globally. Many of their open-source libraries are now gaining popularity on GitHub, which is helping budding AI developers across the world build flexible and scalable machine learning models.

From conversational chatbot, self-driving cars to the weather forecast and recommendation systems, AI developers are experimenting with various neural network architectures, hyperparameters, and other features to fit the hardware constraints of edge platforms. The possibilities are endless. Some of the popular deep learning frameworks include Googles TensorFlow and Facebooks Caffe2, PyTorch, Torchcraft AI and Hydra, etc.

According to Statista, AI business operations global revenue is expected to touch $10.8 billion by 2023, and the natural language processing (NLP) market size globally is expected to reach $43.3 billion by 2025. With the rise of AI adoption across businesses, the need for open-source libraries and architecture will only increase in the coming months.

Advancing in artificial intelligence, Facebook AI Research (FAIR) at present is leading the AI race with the launch of state of the art technology tools, libraries and frameworks to bolster machine learning and AI applications across the globe.

Source: Analytics India Magazine

Here are some of the latest open-source tools, libraries and architecture developed by Facebook:

PyTorch is the most widely used deep learning framework, besides Caffe2 and Hydra, which helps researchers build flexible machine learning models.

PyTorch provides a Python package for high-level features like tensor computation (NumPy) with strong GPU acceleration and TorchScript for an easy transition between eager mode and graph mode. Its latest release provides graph-based execution, distributed training, mobile deployment and more.

Flashlight is an open-source machine learning library that lets users execute AI/ML applications using C++ API. Since it supports research in C++, Flashlight does not need external figures or bindings to perform tasks such as threading, memory mapping, or interoperating with low-level hardware. Thus, making the integration of code fast, direct and straightforward.

Opacus is an open-source high-speed library for training PyTorch models with differential privacy (DP). The library is claimed to be more scalable than existing methods. It supports training with minimal code changes and has little impact on training performance. It also allows the researchers to track the privacy budget expended at any given moment.

PyTorch3D is a highly modular and optimised library that offers efficient, reusable components for 3D computer vision research with the PyTorch framework. It is designed to integrate smoothly with deep learning methods for predicting and manipulating 3D data. As a result, the library can be implemented using PyTorch tensors, handle mini-batches of heterogeneous data, and utilise GPUs for acceleration.

Detectron2 is a next-generation library that provides detection and segmentation algorithms. It is a fusion of Detectron and maskrcnn-benchmark. Currently, it supports several computer vision research work and applications. Detection can be used on Mask R-CNN, RetinaNet, Faster R-CNN, RPN, TensorMask as well.

Detectron is an open-source software architecture that implements object detection algorithms like Mask R-CNN. The software has been written in Python and powered by the Caffe2 deep learning framework.

Detectron has enabled various research project at Facebook, including Feature pyramid networks for object detection, Mask R-CNN, non-local neural networks, detecting and recognising human-object interactions, learning to segment everything, data distillation: towards Omni-supervised learning, focal loss for dense object detection, DensePose: dense human pose estimation in the wild, and others.

Prophet is an open-source architecture released by Facebooks core data science team. It is a procedure for forecasting time series data based on an additive model where non-linear trends fit yearly, weekly, and daily seasonality, plus holiday effects. The model works best with time-series data, which has several seasons of historical data such as weather records, economic indicators and patient health evolution metrics.

The code is available on CRAN and PyPI.

Classy Vision is a new end-to-end PyTorch-based framework for large-scale training of image and video classification models. Unlike other computer vision (CV) libraries, Classy Vision claims to offer flexibility for researchers.

Typically, most CV libraries lead to duplicative efforts and require users to migrate research between frameworks and relearn the minutiae of efficient distributed training and data loading. On the other hand, Facebooks PyTorch-based CV framework claimed to offer a better solution for training at scale and deploying to production.

BoTorch is a library for Bayesian optimization built on the PyTorch framework. Bayesian optimization is a sequence design strategy for machines that do not assume any functional forms.

BoTorch seamlessly provides a modular and easily extensible interface for composing Bayesian optimization primitives such as probabilistic models, acquisition functions and optimizers and others. In addition to this, it also enables seamless integration with deep or convolutional architectures in PyTorch.

FastText is an open-source library for efficient text classification and representation learning. It works on standard and generic hardware. Machine learning models can be further reduced on mobile devices as well.

TC is a fully-functional C++ library that automatically synthesises high-performance machine learning kernels using Halide, ISL, NVRTC or LLVM. The library can be easily integrated with Caffe2 and PyTorch and has been designed to be highly portable and machine-learning framework agnostic. Also, it requires a simple tensor library with memory allocation, offloading, and synchronisation capabilities.

Here is the original post:
All The Machine Learning Libraries Open-Sourced By Facebook Ever - Analytics India Magazine

Application of Immune Infiltration Signature and Machine Learning Model in the Differential Diagnosis and Prognosis of Bone-Related Malignancies -…

This article was originally published here

Front Cell Dev Biol. 2021 Apr 15;9:630355. doi: 10.3389/fcell.2021.630355. eCollection 2021.

ABSTRACT

Bone-related malignancies, such as osteosarcoma, Ewings sarcoma, multiple myeloma, and cancer bone metastases have similar histological context, but they are distinct in origin and biological behavior. We hypothesize that a distinct immune infiltrative microenvironment exists in these four most common malignant bone-associated tumors and can be used for tumor diagnosis and patient prognosis. After sample cleaning, data integration, and batch effect removal, we used 22 publicly available datasets to draw out the tumor immune microenvironment using the ssGSEA algorithm. The diagnostic model was developed using the random forest. Further statistical analysis of the immune microenvironment and clinical data of patients with osteosarcoma and Ewings sarcoma was carried out. The results suggested significant differences in the microenvironment of bone-related tumors, and the diagnostic accuracy of the model was higher than 97%. Also, high infiltration of multiple immune cells in Ewings sarcoma was suggestive of poor patient prognosis. Meanwhile, increased infiltration of macrophages and B cells suggested a better prognosis for patients with osteosarcoma, and effector memory CD8 T cells and type 2 T helper cells correlated with patients chemotherapy responsiveness and tumor metastasis. Our study revealed that the random forest diagnostic model based on immune infiltration can accurately perform the differential diagnosis of bone-related malignancies. The immune microenvironment of osteosarcoma and Ewings sarcoma has an important impact on patient prognosis. Suppressing the highly inflammatory environment of Ewings sarcoma and promoting macrophage and B cell infiltration may have good potential to be a novel adjuvant treatment option for osteosarcoma and Ewings sarcoma.

PMID:33937231 | PMC:PMC8082117 | DOI:10.3389/fcell.2021.630355

See more here:
Application of Immune Infiltration Signature and Machine Learning Model in the Differential Diagnosis and Prognosis of Bone-Related Malignancies -...

YouTube Introduces Machine Learning Individual Object Recognition on Videos and Facebook Follows Next – Digital Information World

Artificial Intelligence is one of the greatest achievements in the tech world and while machine learning was only limited to reading still frames up until now and is quite efficient in it, the next step being taken in the advancement of machine learning and artificial intelligence is identifying individual objects within video in order to open up new considerations in brand placement, visual effects, accessibility features and more.

The first and successful step taken towards making AI identify individual objects within video was done by Google. Google had been working towards accomplishing this feature for some time now and after a lot of efforts it has now introduced new advances in its YouTube option which includes being able to tag products in that are present in video clips and provide direct links to shop for those products.

This simply means that companies now can tag their products in YouTube videos no matter at what timing it is being displayed, it can tag its product at that specific time. Along with this it will also provide direct shopping options, facilitating broader ecommerce opportunities in the app.

After the successful introduction of this feature in YouTube, Facebook is taking the next step and introducing a similar feature on its platform and the company claims that their feature will be much better at singling out individual objects within video frames.

Facebook explained that they have collaborated with researchers at Inria with whom they have developed a new method called DINO. This method will be used to train Vision Transformers (ViT) with no supervision. The company has claimed that besides setting a new state of the art among self-supervised methods, this approach leads to a remarkable result that is unique to this combination of AI techniques. Facebook further said that their model can discover and segment objects in an image or a video with absolutely no supervision and without being given a segmentation-targeted objective and all this will make this process effectively automated.

Hence that is why the company claims that their feature is the best of the best.

Facebook is still working towards this feature and once it is launched we cannot wait to see if it out does YouTubes similar feature or not. However, we know that both YouTube and Facebook have always delivered their best and therefore we are sure that they will deliver the best this time as well.

Read next:According to the exec, over 60 percent Instagram users are connected to Facebook Messenger

Read this article:
YouTube Introduces Machine Learning Individual Object Recognition on Videos and Facebook Follows Next - Digital Information World

AI and Machine Learning Operationalization Software Market by Technology Innovations and Growth 2021 KSU | The Sentinel Newspaper – KSU | The…

The AI & Machine Learning Operationalization Software Market report is a compilation of first-hand information, qualitative and quantitative assessment by industry analysts, inputs from industry experts and industry participants across the value chain. The report provides in-depth analysis of parent market trends, macro-economic indicators and governing factors along with market attractiveness as per segments. The report also maps the qualitative impact of various market factors on market segments and geographies.

Get Sample Copy of this Report:

https://www.marketintelligencedata.com/reports/275731/global-ai-machine-learning-operationalization-software-market-growth-status-and-outlook-2021-2026/inquiry?Mode=RJ

Top LeadingCompaniesof Global AI & Machine Learning Operationalization Software Market areAlgorithmia, Spell, Valohai Ltd, 5Analytics, Cognitivescale, Datatron Technologies, Acusense Technologies, Determined AI, DreamQuark, Logical Clocks, IBM, Imandra, Iterative, Databricks, ParallelM, MLPerf, Neptune Labs, Numericcal, Peltarion, Weights & Biases, WidgetBrainand others.

Regional Outlook of AI & Machine Learning Operationalization Software Market report includes the following geographic areas such as: North America, Europe, China, Japan, Southeast Asia, India and ROW.

On The Basis Of Product, The AI & Machine Learning Operationalization Software Market Is Primarily Split Into

Cloud-Based

Web-Based

On The Basis Of End Users/Application, This Report Covers

Large Enterprises

SMEs

Browse Full Report at:

https://www.marketintelligencedata.com/reports/275731/global-ai-machine-learning-operationalization-software-market-growth-status-and-outlook-2021-2026?Mode=RJ

This allows understanding of the market and benefits from any lucrative opportunities that are available. Researchers have offered a comprehensive study of the existing market scenario while concentrating on the new business objectives. There is a detailed analysis of the change in customer requirements, customer preferences and the vendor landscape of the overall market.

Following are major Table of Content of AI & Machine Learning Operationalization Software Industry:

Furthermore, this study will help our clients solve the following issues:

Cyclical dynamics We foresee dynamics of industries by using core analytical and unconventional market research approaches. Our clients use insights provided by us to maneuver themselves through market uncertainties and interferences.

Identifying key cannibalizes Strong substitute of a product or service is the most important threat. Our clients can identify key cannibalizes of a market, by procuring our research. This helps them in aligning their new product development/launch strategies in advance.

Spotting emerging trends The report help clients to spot upcoming hot market trends. We also track possible impact and disruptions which a market would witness by a particular emerging trend. Our proactive analysis help clients to have early mover advantage.

Interrelated opportunities This report will allow clients to make decisions based on data, thereby increasing the chances that the strategies will perform better if not best in real world.

We Offer Customization on report based on specific client Requirement:

Free country Level analysis for any 5 countries of your choice. Free Competitive analysis of any 5 key market players. Free 40 analyst hours to cover any other data point.

About Us:

Marketintelligencedataprovides syndicated market research on industry verticals includingHealthcare, Information and Communication Technology (ICT), Technology and Media, Chemicals, Materials, Energy, Heavy Industry, etc.Marketintelligencedataprovides global and regional market intelligence coverage, a 360-degree market view which includes statistical forecasts, competitive landscape, detailed segmentation, key trends, and strategic recommendations.

Contact Us:

Irfan Tamboli (Head of Sales) Market Intelligence Data

Phone: + 1704 266 3234 | +91-750-707-8687

sales@marketintelligencedata.com

Go here to read the rest:
AI and Machine Learning Operationalization Software Market by Technology Innovations and Growth 2021 KSU | The Sentinel Newspaper - KSU | The...