Connect with us

Technology

Understanding Graph Neural Networks

Published

on

Understanding Graph Neural Networks

With rich data structures that represent items and their interactions as points connected by lines in a graph, graph neural networks applies predictive potential of deep learning to these types of complex data structures.

Data points in GNNs are referred to as nodes and are connected by lines, referred to as edges, with elements represented mathematically such that machine learning algorithms can make valuable predictions at the level of nodes, edges, or complete graphs.

When applied to rich data structures that represent objects and their interactions as points connected by lines in a graph, GNNs can predict the future with the accuracy of deep learning.

When two technologies come together, something new and beautiful can be produced, much as smartphones were produced when cellphones and browsers combined.

Currently, developers are using enormous graph databases that contain data about relationships among data points of all kinds to use AI’s capacity for pattern recognition. Together, they create graph neural networks, a potent new tool.

GNNs are being used by an increasing number of businesses to enhance recommendation, fraud detection, and drug discovery systems. Finding patterns in the interactions between data items is essential to these applications and many more.

GNN applications in computer graphics, cybersecurity, genomics, and materials science are being investigated by researchers. In a recent research, it was shown how GNNs might forecast arrival times more accurately by using transportation maps as graphs.

Valuable data is already stored in graph databases by many fields of science and business. They can develop predictive models using deep learning to get novel insights from their graphs.

Understanding Graph Neural Networks

Who Are Graph Neural Networks main users?

In order to identify fraud, Amazon deployed Graph Neural Networks in 2017. It launched a free public GNN service in 2020 for applications such as fraud detection, recommendation engines, and other uses.

Amazon Search uses Graph Neural Networks to identify dishonest vendors, buyers, and products in order to uphold their consumers’ high level of confidence. It can explore graphs with tens of millions of nodes and hundreds of millions of edges using NVIDIA GPUs, cutting training time in half from 24 to 5 hours.

The worldwide head of AI for the biopharma business GSK remarked on a panel at a GNN workshop that the company maintains a knowledge graph with approximately 500 billion nodes that is used in several of its machine-language models.

According to Jaewon Yang, a senior staff software engineer at the business, who was participating in a different workshop panel, “LinkedIn employs GNNs to generate social recommendations and identify the correlations between people’s skills and their job titles.”

More than 700,000 people have seen the videos of a GNN course Leskovec gives at Stanford, which is another indication of the interest in GNNs.

Understanding Graph Neural Networks

How Graph Neural Networks Work?

Deep learning has primarily concentrated on text and images up to this point, which are examples of structured data that may be thought of as grids of pixels or word sequences. Graphs, on the other hand, lack structure. They can be any size or shape and can hold any type of data, including text and image data.

GNNs arrange graphs in a manner that machine learning algorithms can use through a procedure known as message forwarding.

Each node receives information about its neighbors through message passing. The embedded data is used by AI models to identify patterns and make predictions.

In addition, GNNs are distinct because they employ sparse mathematics and frequently have just two or three layers in their models. Other AI models typically employ complex mathematics and contain hundreds of layers of neural networks.

The History of GNNs

Graph neural networks were given their name for the first time in a 2009 article by Italian researchers. However, it took eight years for two Amsterdam-based researchers to prove their prowess with a variation they named a graph convolutional network (GCN), which is now among the most widely used GNNs.

Leskovec and two of his Stanford graduate students were inspired by the GCN work to develop GraphSage, a GNN that demonstrated novel applications for the message-passing function. In the summer of 2017, he tested it at Pinterest, where he worked as chief scientist.

For message passing in GNNs, GraphSage invented several effective aggregation approaches.

In order to surpass other AI models at the time, their version, called PinSage, packed 3 billion nodes and 18 billion edges.

More than 100 use cases across the firm of Pinterest currently use it. Andrew Zhai, a senior machine learning engineer at the business, stated on an online panel that “Pinterest would not be as entertaining as it is now without GNNs.”

In the meantime, other variations and hybrids have appeared, such as graph attention networks and recurrent networks. In order to assist GNNs in focusing on the most interesting regions of datasets, GATs take advantage of the attention mechanism outlined in transformer models.

 

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Technology

UK Government To Set Online Bill Criminalizing Self Harm

Published

on

UK Government Sets Online Bill To Criminalize Self Harm

In an effort to stop what it calls “tragic and preventable deaths caused by people seeing self-harm content online,” the UK government has announced it will further broaden the scope of online safety legislation by making encouraging self-harm a crime.

According to the most recent modification to the divisive but popular Online Safety Bill, in-scope platforms would be compelled to remove anything that purposefully encourages someone to physically harm themselves, or face legal repercussions.

The government intends to tackle “abhorrent trolls urging the young and vulnerable to self-harm,” according to the secretary of state for digital. People who post such content online may also be prosecuted under the new offence of encouraging self injury.

The maximum fines will be announced in due time, according to the administration.

In the UK, it is already unlawful to promote or aid suicide, whether in person or online. By creating a new offense, self-harm content will now be subject to the same laws that already ban suicide promotion.

Following a snag, last summer associated with political unrest in the ruling Conservative Party, the Online Safety Bill’s progress through parliament is now on hold. However, the newly reorganized UK government has declared that it will reintroduce the measure to parliament next month after making changes to the law.

The abuse of intimate imagery is a problem that will be addressed by recent revisions to the Online Safety Bill, which was just made public by the Ministry of Justice. However, other revisions are planned regarding “legal but harmful” information, thus the final form of the Act is still up in the air.

The government responded to concerns about the bill’s impact on online freedom of expression a few months ago. The (new) secretary of state, Michelle Donelan, announced in September that she would be “editing” the bill to lessen concern about its impact on “legal but harmful” speech for adults.

The most recent changes, making it illegal to send online communications encouraging self harm, came after that announcement.

UK Government Sets Online Bill To Criminalize Self Harm

Donelan was quoted by the BBC as claiming that Molly Russell, a 14-year-old teenager who committed suicide five years ago after watching thousands of online articles on self-harm and suicide on websites like Instagram and Pinterest, was a factor in the most recent changes.

Social media was found to have contributed to Russell’s death, according to the results of an inquest into her death in September. While the coroner’s “prevention of future deaths” report from last month that a number of steps be done to control and monitor young people’s access to social media content.

The addition of the crime of promoting self harm, according to the Department for Digital, Culture, Media, and Sport, will outlaw “one of the most worrying and prevalent internet harms that now falls below the threshold of criminal behavior.”

Donelan stated in a statement:

“I am determined that the abhorrent trolls encouraging the young and vulnerable to self-harm are brought to justice.

“So I am strengthening our online safety laws to make sure these vile acts are stamped out and the perpetrators face jail time.

“Social media firms can no longer remain silent bystanders either and they’ll face fines for allowing this abusive and destructive behaviour to continue on their platforms under our laws.”

Hate crimes, rules regarding revenge porn (including disseminating deepfake porn without content), harassment, and cyberstalking are among the other top criminal offenses already mentioned in the bill.

Regardless of what the measure states on paper, there are still a lot of unknowns regarding how platforms will react to having legal obligations imposed on them to police all forms of speech, as well as if it would actually increase web user safety as claimed.

Critics worry that the regime will have a chilling effect by turning platforms into de facto speech police and encouraging them to overblock content in order to reduce their legal risk of paying a hefty fine.

The regime’s penalties scale up to 10% of global annual turnover, and non-cooperative senior executives even run the risk of going to jail.

On Monday, December 5, the bill is scheduled to return to parliament.

 

 

 

 

 

 

 

Continue Reading

Technology

Twitter Amnesty Is What Elon Musk is Going For Next

Published

on

Twitter Amnesty Is What Elon Musk is Going For Next

Tesla CEO and newly appointed Twitter CEO, Elon Musk did promise a new dimension for the micro-blogging social media platform prior to taking over, and his actions recently, have just about lived up to the promise, but now, the billionaire is set for an ‘amnesty’ that surely will drive some political divides nuts if certain individuals are granted Twitter amnesty as he wants.

Elon Musk announced on Thursday that starting the next week, Twitter will provide suspended accounts “a general amnesty.” The day before, the platform’s CEO published a poll asking users if they thought affected accounts should be restored.

The announcement comes just after Musk lifted the platform’s restriction on former president Donald Trump after conducting a related poll. Trump declared he had no intention of returning to the platform despite being banned following the attack on the US Capitol on January 6, 2021.

Users of the Twitter platform who had their accounts suspended could rejoin the network “assuming they have not broken the law or engaged in egregious spam,” according to Musk’s user survey.

Twitter Amnesty Is What Elon Musk is Going For Next

The survey received responses from about 3.2 million individuals, who voted 72.4% in favor of amnesty.

“The people have spoken. Amnesty begins next week. Vox Populi, Vox Dei,” Musk said, using a Latin phrase that means “The voice of the people is the voice of god.”

Historically, Twitter has deactivated accounts who advocate violence, celebrate hate and harassment, or persistently disseminate false information that may be harmful.

Some well-known people who were banned from the website include MyPillow CEO Mike Lindell, who made a number of claims that Trump actually won the 2020 presidential election, former Trump advisor and former executive chairman of Breitbart Steve Bannon, who said Anthony Fauci and FBI Director Christopher Wray should be beheaded, and Proud Boys founder Gavin McInnes, who broke the website’s rule against violent extremist groups.

Considering that more voices with possibly negative views will be returning to the site, it’s unclear from Musk’s brief post how Twitter will handle content control going forward.

These worries have only grown as a result of Musk’s huge firings and the outflow of workers who would rather leave than remain “hardcore.”

Elon Musk is surely growing more unpopular by remaining popular these days.

Continue Reading

Technology

Twitter Working On New Feature For Long Texts

Published

on

Twitter Working On New Feature For Long Texts

Writing a thread on Twitter can be considered daunting especially when you have to divide the text into 280-character sections for it to make meaning.

Good news though as the platform is stated to be working on a way to convert lengthy texts into threads automatically.

When a tweet exceeds the 280-character limit, Twitter’s composer will automatically divide it into a thread, according to a tweet from app researcher Jane Manchun Wong.

Twitter wants to make making threads less difficult, as she stated in a message to a user (identified as me).

Currently, in order to add a tweet to a thread and post the subsequent 280 characters, users must click the Add button. This can be particularly unpleasant when you are trying out an idea or pasting information from another document.

Several users have recently brought up the difficulty posting to and reading conversations with more than a few tweets; the thread in question was 82 tweets long and focused on the defunct crypto-currency exchange FTX. In response, Musk stated that the team is working to make thread writing simpler.

It will be useful to have markers to designate the start and end of a tweet in the thread, although the exact implementation details remain unknown, as Financial Times product manager Matt Taylor noted. This makes it simpler for users to change the text in a way that doesn’t disrupt the reading flow.

Musk has previously addressed the problem of posting lengthy tweets. He previously stated that the social network is developing the capability to attach long-form content to tweets. If that will be a standalone feature from the new thread composer is unclear.

Currently, some users rely on third-party programs like Typefully, ThreadStart, and Chirr App, which offer capabilities like scheduling along with tools to automatically divide your post into threads without interfering with sentence flow.

Thanks to its acquisition of Threader the previous year, the company today provides Twitter Blue customers with a simple way to read threads. However, Musk hasn’t actually stated whether he is altering the reading experience for the typical user.

There is already a long-form writing program on Twitter called Notes, but it is exclusively available to a small number of writers, and under Musk’s leadership, its future is unclear.

Even though Twitter programmers are already working on it, it is unclear when the new composer tool for threads will launch. Since taking over the business, Musk has let go of more than half the employees.

Numerous executives have left, and the new leader even gave the remaining employees yesterday an ultimatum: either be “hardcore” or quit. There is no assurance that goods will be delivered on time in this situation.

The new Twitter Blue plan with a verification mark was hurriedly launched by the firm, only for the scheme to be discontinued a few days later. Musk stated earlier this week that the launch date had been moved to later in the month.

Wong just found code that suggests Twitter is working on encrypting direct communications from end to end.

 

Continue Reading

Trending