Simulation noise is a function that creates a divergence-free vector field. This signal can be used in artistic simulations for the purpose of increasing the perception of extra detail. The function can be calculated in three dimensions by dividing the space into a regular lattice grid. With each edge is associated a random value, indicating a rotational component of material revolving around the edge. By following rotating material into and out of faces, one can quickly sum the flux passing through each face of the lattice. Flux values at lattice faces are then interpolated to create a field value for all positions. Perlin noise is the earliest form of lattice noise, which has become very popular in computer graphics. Perlin Noise is not suited for simulation because it is not divergence-free. Noises based on lattices, such as simulation noise and Perlin noise, are often calculated at different frequencies and summed together to form band-limited fractal signals. Other approaches developed later that use vector calculus identities to produce divergence free fields, such as "Curl-Noise" as suggested by Rook Bridson, and "Divergence-Free Noise" due to Ivan DeWolf. These often require calculation of lattice noise gradients, which sometimes are not readily available. A naive implementation would call a lattice noise function several times to calculate its gradient, resulting in more computation than is strictly necessary. Unlike these noises, simulation noise has a geometric rationale in addition to its mathematical properties. It simulates vortices scattered in space, to produce its pleasing aesthetic. == Curl noise == The vector field is created as follows, for every point (x,y,z) in the space a vector field G is created, every component x, y and z of the vector field (Gx, Gy, Gz) is defined by a 3D perlin or simplex noise function with x, y and z as parameters. The partial derivative of Gx, Gy, and Gz respect to x, y and z is obtained with the gradient of the perlin or simplex noise by finite differences of implicit calculation inside the simplex noise. The partial derivatives are used to calculate F as the curl of G given by F = ( ∂ G z ∂ y − ∂ G y ∂ z , ∂ G x ∂ z − ∂ G z ∂ x , ∂ G y ∂ x − ∂ G x ∂ y ) {\displaystyle F=({\frac {\partial Gz}{\partial y}}-{\frac {\partial Gy}{\partial z}},{\frac {\partial Gx}{\partial z}}-{\frac {\partial Gz}{\partial x}},{\frac {\partial Gy}{\partial x}}-{\frac {\partial Gx}{\partial y}})} == Bitangent noise == This method is based in the fact that the curl of the gradient of scalar field is zero and the identity that expand the divergence of a cross product of two vectors A and B as the difference of the dot products of each vector with the curl of the other: ∇ × ( ∇ φ ) = 0 . {\displaystyle \nabla \times (\nabla \varphi )=\mathbf {0} .} ∇ ⋅ ( A × B ) = ( ∇ × A ) ⋅ B − A ⋅ ( ∇ × B ) {\displaystyle \nabla \cdot (\mathbf {A} \times \mathbf {B} )=\ (\nabla {\times }\mathbf {A} )\cdot \mathbf {B} \,-\,\mathbf {A} \cdot (\nabla {\times }\mathbf {B} )} which means that if the curl of both vector fields is zero then the divergence of the product of two vectors that are the gradients of scalar fields is zero too. This result in a divergence free vector field by construction only calling two noise functions to create the scalar fields. The vector field es created as follows, two scalar fields are calculated ϕ {\displaystyle \phi } and ψ {\displaystyle \psi } using 3D perlin or simplex noise functions, then the gradients A and B of each of this fields is calculated, the cross product of A and B gives a divergence free vector field. == Signed distance noise == The vector field is created based on a closed and differentiable implicit surface S = F(x,y,z) = 0. For every point in the space, frequently outside or near the surface, we get a vector g that is normal to the surface, this is the gradient of S or the partial derivatives respect to x, y and z, this vector is not unitary, but we can get a unitary normal n by dividing each component of the point by the magnitude of the gradient g. Outside of the surface all these normals point away from the surface. g = ∇ F ( x , y , z ) = ( ∂ F ∂ x , ∂ F ∂ y , ∂ F ∂ z ) {\displaystyle g=\nabla F(x,y,z)=\left({\frac {\partial F}{\partial x}},{\frac {\partial F}{\partial y}},{\frac {\partial F}{\partial z}}\right)} n = g ( x , y , z ) ‖ ∇ F ( x , y , z ) ‖ {\displaystyle \mathbf {n} ={\frac {g(x,y,z)}{\|\nabla F(x,y,z)\|}}} ‖ ∇ F ( x , y , z ) ‖ = ( ∂ F ∂ x ) 2 + ( ∂ F ∂ y ) 2 + ( ∂ F ∂ z ) 2 {\displaystyle \|\nabla F(x,y,z)\|={\sqrt {\left({\frac {\partial F}{\partial x}}\right)^{2}+\left({\frac {\partial F}{\partial y}}\right)^{2}+\left({\frac {\partial F}{\partial z}}\right)^{2}}}} Afterwards we calculate a scalar value p for that point in the space using a 3D perlin or simplex noise function. Now we create a vector field V = pn pointing outside of the surface. The curl of this vector field gives the direction in every point in the space where the particles should move. S D N = ( ∂ V z ∂ y − ∂ V y ∂ z , ∂ V x ∂ z − ∂ V z ∂ x , ∂ V y ∂ x − ∂ V x ∂ y ) {\displaystyle SDN=({\frac {\partial Vz}{\partial y}}-{\frac {\partial Vy}{\partial z}},{\frac {\partial Vx}{\partial z}}-{\frac {\partial Vz}{\partial x}},{\frac {\partial Vy}{\partial x}}-{\frac {\partial Vx}{\partial y}})} By construction this vector SDN will point in a tangent direction to an isosurface at the level of the signed distance to the original surface and can be used to confine the movements of the particles to stay in that surface.
Shell Control Box
Shell Control Box (SCB) is a network security appliance that controls privileged access to remote IT systems, records activities in replayable audit trails, and prevents malicious actions. For example, it records as a system administrator updates a file server or a third-party network operator configures a router. The recorded audit trails can be replayed like a movie to review the events as they occurred. The content of the audit trails is indexed to make searching for events and automatic reporting possible. SCB is a Linux-based device developed by Balabit. It is an application level proxy gateway. In 2017, Balabit changed the name of the product to Privileged Session Management (PSM) and repositioned it as the core module of its Privileged Access Management solution. == Main Features == Balabit’s Privileged Session Management (PSM), Shell Control Box (SCB) is a device that controls, monitors, and audits remote administrative access to servers and network devices. It is a tool to oversee system administrators by controlling the encrypted connections used for administration. PSM (SCB) has full control over the SSH, RDP, Telnet, TN3270, TN5250, Citrix ICA, and VNC connections, providing a framework (with solid boundaries) for the work of the administrators. === Gateway Authentication === PSM (SCB) acts as an authentication gateway, enforcing strong authentication before users access IT assets. PSM can also integrate to user directories (for example, a Microsoft Active Directory) to resolve the group memberships of the users who access the protected servers. Credentials for accessing the server are retrieved transparently from PSM’s credential store or a third-party password management system by PSM impersonating the authenticated user. This automatic password retrieval protects the confidentiality of passwords as users can never access them. === Access Control === PSM controls and audits privileged access over the most wide-spread protocols such as SSH, RDP, or HTTP(s). The detailed access management helps to control who can access what and when on servers. It is also possible to control advanced features of the protocols, like the type of channels permitted. For example, unneeded channels like file transfer or file sharing can be disabled, reducing the security risk on the server. With PSM policies for privileged access can be enforced in one single system. === 4-eyes Authorization === To avoid accidental misconfiguration and other human errors, PSM supports the 4-eyes authorization principle. This is achieved by requiring an authorizer to allow administrators to access the server. The authorizer also has the possibility to monitor – and terminate - the session of the administrator in real-time, as if they were watching the same screen. === Real-time Monitoring and Session Termination === PSM can monitor the network traffic in real time, and execute various actions if a certain pattern (for example, a suspicious command, window title or text) appears on the screen. PSM can also detect specific patterns such as credit card numbers. In case of detecting a suspicious user action, PSM can send an e-mail alert or immediately terminate the connection. For example, PSM can block the connection before a destructive administrator command, such as the „rm” comes into effect. === Session Recording === PSM makes user activities traceable by recording them in tamper-proof and confidential audit trails. It records the selected sessions into encrypted, timestamped, and digitally signed audit trails. Audit trails can be browsed online, or followed real-time to monitor the activities of the users. PSM replays the recorded sessions just like a movie – actions of the users can be seen exactly as they appeared on their monitor. The Balabit Desktop Player enables fast forwarding during replays, searching for events (for example, typed commands or pressing Enter) and texts seen by the user. In the case of any problems (database manipulation, unexpected shutdown, etc.) the circumstances of the event are readily available in the trails, thus the cause of the incident can be identified. In addition to recording audit trails, transferred files can be also recorded and extracted for further analysis.
IAmAnas
#IAmAnas (I Am Anas) is a Twitter hashtag and social media campaign that started in 2015. Users tweeted to express support for the undercover investigative works of Ghanaian journalist Anas Aremeyaw Anas. The campaign restarted in 2018 when the Ghanaian MP and financier of the New Patriotic Party, Kennedy Agyapong, announced his intention to reveal the identity of Anas following the journalist's exposé of corruption at the Ghana Football Association. Anas maintains that "being anonymous has always been his secret weapon." Pictures purported to be of Anas were first released by a TV station owned by Agyapong, and were quickly picked up by other media houses. At least one person, a Dutch-Brazilian model, has claimed ownership of one picture that was released, and has threatened legal action against Agyapong for possibly putting his life in danger. In response to Agyapong, social media users retweeted photos of themselves, random people, or even comic images of entities that resemble the trademark covered face of Anas. When the hashtag first began in 2015, along with other popular uses of the journalist's name, Elizabeth Ohene wrote an article about Ghanaians use of humour in response to dealing with the expose of government corruption. "I do not know when these words will make it into Wikipedia or the Oxford English Dictionary but for the moment you can take it from me that: To go undercover is to anas, to make secret recordings is to anas-anas, to wear disguises is to do an anas, to be caught in the act is to be anased. To have someone exposed taking bribes is to have that person being given the full Anas Aremeyaw Anas."
Groundswell (book)
Groundswell is a book by Forrester Research executives Charlene Li and Josh Bernoff that focuses on how companies can take advantage of emerging social technologies. It was published in 2008 by Harvard Business Press. A revised edition was published in 2011. The book attempts to explain a shift in the relationship between customers and companies, in which companies are no longer able to control customers' attitudes through market research, customer service, and advertising. Instead, customers are controlling the conversation by using new media to communicate about products and companies. == Synopsis == The groundswell is characterized by several tactics that guide companies into using social technologies strategically and effectively. Listening: Businesses should listen to their customers to understand what the market is looking for in their products. In order to do this, a company needs to find out if their customers are using social technologies and how they are using them. Talking: Instead of advertising to customers, marketing departments should find creative ways to connect with users about their experience with a product and their feelings about the brand. One common method is participation in social networks. Energizing: Enthusiastic customers are part of the groundswell, and companies can recognize and appreciate these customers by creating online communities and social platforms where they can connect with the brand and provide reviews. Supporting: Businesses can harness the support of their own employees by creating internal social applications for them to connect with the brand, also known as enterprise social software. == Groundswell in action == === Examples === Some companies distinguish their product through the use of social technologies. Tom Dickson successfully marketed his Blendtec line of blenders through the viral marketing campaign Will It Blend? The groundswell spread marketing messages through Digg and YouTube with a small budget and little marketing experience. Other companies have been able to listen to and talk with the groundswell by building their own online communities. Procter & Gamble created beinggirl.com Archived 2016-04-10 at the Wayback Machine to introduce girls to P&G feminine care products. The community approach worked because the company could reach girls with information that might seem embarrassing or sensitive in a traditional marketing campaign. === Risks === Features of particular industries or companies can make direct customer engagement more difficult. For instance, some companies must work within industry regulations, national or multinational corporations must balance corporate and local engagement, and other companies must find ways to engage with customers on time-sensitive issues. == Reception == Kevin Allison of the Financial Times praised the book for its focus on Web analytics: "[Groundswell] is not so much a manifesto or a dissection of online culture as it is a how-to manual for executives and mid-level managers trying to navigate this fast-changing and often confusing environment." The book won the American Marketing Association Foundation’s Berry-AMA Book Prize for best marketing book of 2009. It was also listed by: Amazon, as one of the Top 10 Business & Investing Books of 2008 CIO Insight, as one of the Top 10 Business-Tech Books of 2008 and one of 10 Insightful Web 2.0 Books Fortune as Magazine as one of the 3 best Web books of 2008 Advertising Age as number 3 of 10 Books You Should Have Read BusinessWeek as one of the Best Innovation & Design Books of 2008 "strategy+business" as one of the Best Business Books 2008 and “Top Shelf” in Marketing
G7 Rapid Response Mechanism
The G7 Rapid Response Mechanism (RRM) is an initiative introduced in the "Charlevoix Commitment on Defending Democracy from Foreign Threats", issued by the leaders of the Group of Seven (G7) countries—United States, Canada, Japan, United Kingdom, France, Germany and Italy—on June 9, 2018, during their summit in Charlevoix, Quebec. The RRM's mandate is to strengthen the coordination of G7 member countries, as well as "to identify and respond to diverse and evolving threats to our democracies, including through sharing information and analysis, and identifying opportunities for coordinated response" The G7 is an informal international intergovernmental economic organization that meets annually, whose members represent the seven wealthiest advanced economies in the world, as measured by the International Monetary Fund (IMF). == Constituents == The following countries and organisations are members and observers (associate members) of the G7 Rapid Response Mechanism: Australia Canada France Germany Italy Japan Netherlands New Zealand Poland Sweden United Kingdom United States European Union North Atlantic Treaty Organization == Mandate == The RRM was mandated to "strengthen coordination to prevent, thwart and respond to malign and evolving threats to G7 democracies." It "will share information and threat analysis related to various threats to democracy, and is an established mechanism to identify opportunities for coordinated response." According to the Institute for Research on Public Policy's Policy Options magazine, the "RRM initiative seeks to strengthen the leading democracies' coordination to identify and respond to diverse and evolving threats…including through sharing information and analysis, and identifying opportunities for a coordinated response." == Administration == The RRM initiative is led by Canada through Global Affairs Canada's Centre for International Digital Policy. Tara Denham, Director of the Centre for International Digital Policy at Global Affairs Canada, directed the team responsible for setting up the RRM Coordination Unit. Global Affairs Canada—the Department of Foreign Affairs, Trade and Development—is the federal Canadian ministry responsible for diplomatic and consular relations, international trade, and international development and humanitarian assistance. The Centre for International Digital Policy includes the Digital Inclusion Lab and the RRM. Denham is also the RRM's Canadian Focal Point. At a briefing on "the security and intelligence threats to elections" of the House of Commons Standing Committee on Access to Information, Privacy and Ethics, the chair Bob Zimmer (CPC), said that the 2019 general election "may be different" from past elections in Canada. as the "tools that were used to strengthen civic engagement are being used to undermine, disrupt and destabilize democracy." "Democracies around the world have entered a new era—an era of heightened threat and heightened vigilance—and 2019 will see a number of countries brace for volleys of attempted disruption: India, Australia, Ukraine, Switzerland, Belgium, the EU and, of course, Canada. Evidence has confirmed that the most recent Canadian general election, in 2015, was unencumbered by interference, although there were some relatively primitive attempts to disrupt, misinform and divide. These efforts were few in number and uncoordinated, and had no visible impact on the voter, either online or in line." Zimmer described the initiative's three pillars. "enhancing citizen preparedness" through the "digital citizen initiative" "improving organizational readiness" with national security and intelligence agencies supporting Elections Canada "ensure a comprehensive understanding of and response to any threats to Canada's democratic process." by establishing the Security and Intelligence Threats to Elections Task Force (SITE) which works as a team with the Communications Security Establishment (CSE), the Canadian Security Intelligence Service (CSIS), the Royal Canadian Mounted Police (RCMP), as well as Global Affairs Canada Zimmer said that as part of the third pillar, "We have activated the G7 rapid response mechanism, announced at the G7 leaders' summit in Charlevoix, to strengthen coordination among our G7 allies and to ensure that there is international collaboration and coordination in responding to foreign threats to democracy." == Background == === Charlevoix summit === The G7 met from June 8 to 9, 2018 during their summit at the Manoir Richelieu in Charlevoix, in La Malbaie, Quebec. The Charlevoix Summit was the 44th G7 summit. The group issued eight "Commitments" at the summit. They included: Commitment on Defending Democracy from Foreign Threats Commitment on Equality and Economic Growth Commitment to End Sexual and Gender-Based Violence, Abuse and Harassment in Digital Contexts Declaration on Quality Education for Girls, Adolescent Girls and Women in Developing Countries Commitment on Innovative Financing for Development. Prime Minister Justin Trudeau announced five themes for Canada's G7 presidency which began in January 2018. === Defending Democracy from Foreign Threats === "We commit to take concerted action in responding to foreign actors who seek to undermine our democratic societies and institutions, our electoral processes, our sovereignty and our security as outlined in the Charlevoix Commitment on Defending Democracy from Foreign Threats. We recognize that such threats, particularly those originating from state actors, are not just threats to G7 nations, but to international peace and security and the rules-based international order. We call on others to join us in addressing these growing threats by increasing the resilience and security of our institutions, economies and societies, and by taking concerted action to identify and hold to account those who would do us harm." They committed to "cooperate in defending democracies from foreign threats and establish a response mechanism for that purpose". "Democracy and the rules-based international order are increasingly being challenged by authoritarianism and the defiance of international norms. In particular, foreign actors seek to undermine our democratic societies and institutions, our electoral processes, our sovereignty and our security. These malicious, multi-faceted and ever-evolving tactics constitute a serious strategic threat which we commit to confront together, working with other governments that share our democratic values." The Charlevoix Commitment states that "foreign actors seek to undermine our democratic societies and institutions, our electoral processes, our sovereignty and our security. These malicious, multi-faceted and ever-evolving tactics constitute a serious strategic threat which we commit to confront together, working together with other governments that share our democratic values." The Charlevoix Summit resolved to "establish a G7 Rapid Response Mechanism to strengthen our coordination to identify and respond to diverse and evolving threats to our democracies, including through sharing information and analysis, and identifying opportunities for coordinated response." == Monitored elections == === 2019 European Parliament election === RRM Canada's comprehensive report on the 2019 European Parliament election analyzed open data "related to foreign interference during and leading up to the 2019 European Union Parliamentary Elections, May 23–26, 2019". RRM Canada did not find "significant evidence of state-based foreign interference, or any large-scale, organized and coordinated efforts by non-state actors". They did find that "national or international non-state actors" used tactics based on those used by the Russian sponsored Internet Research Agency (IRA) in previous elections, "such as the 2016 U.S. Elections". For example, blogs, webpages, and social media accounts on Twitter, Facebook and Reddit "were used to spread divisive and false information to damage and negatively impact social cohesion and trust in democratic processes and institutions" in coordinated networks of Facebook groups. === 2019 Alberta general election === RRM Canada's analyz report on the 2019 Alberta general election was intended to "identify any emerging tactics in foreign interference and draw lessons learned for the Canadian general elections scheduled to take place in October 2019." No foreign activity was detected, although the data revealed ""suspicious account creation pattern that is indicative of troll or bot activity". They found "automated inauthentic behaviour and trolling activities" but concluded that they were "very likely domestic". The data showed "suspicious account creation pattern that is indicative of troll or bot activity", and "spikes in account creation" which suggested the "presence of accounts developed for a specific purpose." The accounts were very likely domestic and were "mainly comprised of supporters of the United Conservative Party (UCP)." A seco
Self-supervised learning
Self-supervised learning (SSL) is a paradigm in machine learning where a model is trained on a task using the data itself to generate supervisory signals, rather than relying on externally-provided labels. In the context of neural networks, self-supervised learning aims to leverage inherent structures or relationships within the input data to create meaningful training signals. SSL tasks are designed so that solving them requires capturing essential features or relationships in the data. The input data is typically augmented or transformed in a way that creates pairs of related samples, where one sample serves as the input, and the other is used to formulate the supervisory signal. This augmentation can involve introducing noise, cropping, rotation, or other transformations. Self-supervised learning more closely imitates the way humans learn to classify objects. During SSL, the model learns in two steps. First, the task is solved based on an auxiliary or pretext classification task using pseudo-labels, which help to initialize the model parameters. Next, the actual task is performed with supervised or unsupervised learning. Self-supervised learning has produced promising results in recent years, and has found practical application in fields such as audio processing, and is being used by Facebook and others for speech recognition. == Pseudo-labels == Pseudo-labels are automatically generated labels that a model assigns to unlabeled data based on its own predictions. They are widely used in self-supervised and semi-supervised learning, where ground-truth annotations are limited or unavailable. By treating predicted labels as surrogate ground truth, learning algorithms can make use of large quantities of unlabeled data in the training process. Pseudo-labeling also plays an important role in systems that must adapt to concept drift, where the statistical properties of the data change over time. In these scenarios, the model may detect that an incoming instance deviates from previously learned behavior. The system then generates a classification result for that instance, and this predicted class is used as a pseudo-label for updating or retraining model components that are becoming outdated. This approach enables continuous adaptation in dynamic environments without requiring manual annotation. In many adaptive learning pipelines, pseudo-labels are chosen when the classifier produces sufficiently confident predictions, reducing the risk of propagating errors. These pseudo-labeled instances are then incorporated into training to refresh or evolve the model's understanding of emerging data patterns, particularly when existing components show signs of “aging” due to drift or distributional shifts. This strategy reduces reliance on manual labeling while helping maintain long-term model performance. == Types == === Autoassociative self-supervised learning === Autoassociative self-supervised learning is a specific category of self-supervised learning where a neural network is trained to reproduce or reconstruct its own input data. In other words, the model is tasked with learning a representation of the data that captures its essential features or structure, allowing it to regenerate the original input. The term "autoassociative" comes from the fact that the model is essentially associating the input data with itself. This is often achieved using autoencoders, which are a type of neural network architecture used for representation learning. Autoencoders consist of an encoder network that maps the input data to a lower-dimensional representation (latent space), and a decoder network that reconstructs the input from this representation. The training process involves presenting the model with input data and requiring it to reconstruct the same data as closely as possible. The loss function used during training typically penalizes the difference between the original input and the reconstructed output (e.g. mean squared error). By minimizing this reconstruction error, the autoencoder learns a meaningful representation of the data in its latent space. === Contrastive self-supervised learning === For a binary classification task, training data can be divided into positive examples and negative examples. Positive examples are those that match the target. For example, if training a classifier to identify birds, the positive training data would include images that contain birds. Negative examples would be images that do not. Contrastive self-supervised learning uses both positive and negative examples. The loss function in contrastive learning is used to minimize the distance between positive sample pairs, while maximizing the distance between negative sample pairs. An early example uses a pair of 1-dimensional convolutional neural networks to process a pair of images and maximize their agreement. Contrastive Language-Image Pre-training (CLIP) allows joint pretraining of a text encoder and an image encoder, such that a matching image-text pair have image encoding vector and text encoding vector that span a small angle (having a large cosine similarity). InfoNCE (Noise-Contrastive Estimation) is a method to optimize two models jointly, based on Noise Contrastive Estimation (NCE). Given a set X = { x 1 , … x N } {\displaystyle X=\left\{x_{1},\ldots x_{N}\right\}} of N {\displaystyle N} random samples containing one positive sample from p ( x t + k ∣ c t ) {\displaystyle p\left(x_{t+k}\mid c_{t}\right)} and N − 1 {\displaystyle N-1} negative samples from the 'proposal' distribution p ( x t + k ) {\displaystyle p\left(x_{t+k}\right)} , it minimizes the following loss function: L N = − E X [ log f k ( x t + k , c t ) ∑ x j ∈ X f k ( x j , c t ) ] {\displaystyle {\mathcal {L}}_{\mathrm {N} }=-\mathbb {E} _{X}\left[\log {\frac {f_{k}\left(x_{t+k},c_{t}\right)}{\sum _{x_{j}\in X}f_{k}\left(x_{j},c_{t}\right)}}\right]} === Non-contrastive self-supervised learning === Non-contrastive self-supervised learning (NCSSL) uses only positive examples. Counterintuitively, NCSSL converges on a useful local minimum rather than reaching a trivial solution, with zero loss. For the example of binary classification, it would trivially learn to classify each example as positive. Effective NCSSL requires an extra predictor on the online side that does not back-propagate on the target side. === Joint-Embedding and Predictive Architectures === A major class of self-supervised learning moves beyond contrastive pairs, instead maximizing the agreement between views while preventing collapse through statistical constraints. Rooted in Deep Canonical Correlation Analysis (Deep CCA), this approach includes Joint-Embedding Architectures (JEA) like Barlow Twins and VICReg, which enforce covariance constraints to learn invariant representations without negative sampling. Deep Latent Variable Path Modelling (DLVPM) generalizes this to multimodal systems, using path models to enforce correlation and orthogonality across diverse data types. In 2022 Yann LeCun introduced Joint-Embedding Predictive Architectures (JEPA) as a step towards decision making, reasoning, and autonomous human intelligence in machines, including self-improvement through autonomous learning. Founded in representation learning, LeCun included the concept of a “world model” in JEPA which aims to enable machines to replicate human intellect by providing machines with a concept for the world in which they exist. Unlike autoencoders, JEPAs operate entirely in latent space, avoiding pixel-level noise to focus on semantic structure. Rather than just learning invariance, JEPAs learn by predicting masked latent representations from visible context. JEPA has been applied to domains such as image analysis, audio processing, and motion in images and video. == Comparison with other forms of machine learning == SSL belongs to supervised learning methods insofar as the goal is to generate a classified output from the input. At the same time, however, it does not require the explicit use of labeled input-output pairs. Instead, correlations, metadata embedded in the data, or domain knowledge present in the input are implicitly and autonomously extracted from the data. These supervisory signals, extracted from the data, can then be used for training. SSL is similar to unsupervised learning in that it does not require labels in the sample data. Unlike unsupervised learning, however, learning is not done using inherent data structures. Semi-supervised learning combines supervised and unsupervised learning, requiring only a small portion of the learning data be labeled. In transfer learning, a model designed for one task is reused on a different task. Training an autoencoder intrinsically constitutes a self-supervised process, because the output pattern needs to become an optimal reconstruction of the input pattern itself. However, in current jargon, the term 'self-supervised' often refers to tasks based on a pretext-task training setup
MX1 Ltd
MX1 was a global media services provider founded in July 2016 from a merger between digital media services companies, RR Media and SES Platform Services, and a wholly owned subsidiary of global satellite owner and operator, SES. In September 2019, MX1 was merged into the SES Video division and the MX1 brand dropped. Broadcast and streamed content management, playout, distribution, and monetisation services from both MX1 and SES Video are now provided under the SES name. Before merger with SES, MX1 claimed to manage more than 5 million media assets and every day to distribute more than 3,600 TV channels, manage the playout of over 525 channels, distribute content to more than 120 subscription VOD platforms, and deliver over 8,400 hours of online video streaming and more than 620 hours of premium sports and live events. == Services == MX1 video and media services are provided through a single hybrid, cloud and on-premises solution, called MX1 360, which enables video and media solutions including content and metadata management, archiving, localisation solutions, channel playout, VOD, online video (OTT) and content distribution. Services provided by MX1 include: === Content aggregation === Acquisition of content via satellite, fibre or IP with satellite downlinking services (for encryption, re-encryption and re-muxing into different platforms), fibre reception from any location, and IP reception via the public Internet. Live sports, news and entertainment production (including in-studio, outside broadcasting, and SNG) with mobile live streaming and video contribution. === Content management === Digital mastering including scanning, conversion, restoration, quality control and localisation/versioning. Content archiving including secure, cloud and on-premises digital storage, and disaster recovery services. Metadata packaging and platform validation to enhance content discovery, searchability and cataloguing. Playout preparation and delivery to any format. === Channel origination and playout === Managed TV channel origination in SD, HD and UHD including 3D graphics, and video and audio effects, using cloud-based solution accessible from any location, with live content insertion and operation, and 24/7 monitoring. === Online video/VOD services === Content preparation and management for online video, VOD, live streaming services and Online video platforms using an ultra-high capacity content delivery network, including subscriber management, apps, DRM, social media, advertising tools, monetisation tools, metadata management, and analytics. === Content delivery === Delivery in all video formats over hybrid distribution network of satellite (using over 150 platforms), fibre (60 digital media hubs worldwide) and the Internet with complete downlink/uplink turnaround services and OTT content delivery. == Locations == MX1 has 16 offices worldwide, the most recent opened in March 2017 in Seoul, South Korea, as well as media centres in UK (London), US (Hawley, PA), Israel (Emeq Ha'Ela), Romania (Bucharest) and at the headquarters in Unterföhring near Munich, Germany. In the early part of 2017, significant upgrades were made to MX1's US media centre in Hawley, Pennsylvania, including expanding its capabilities for US based and global content aggregation, management and delivery to support US broadcasters and content providers. == History == RRsat was founded in Israel by David Rivel, an electronics, computers and communications engineer in 1981 as a communications provider, and in 2014 changed its name to RR Media to reflect its expanding global service offering. In 2015, RR Media acquired Eastern Space Systems (ESS), a Romanian provider of content management and content distribution services and satellite transmission services provider, SatLink Communications. Digital Playout Centre GmbH (DPC) was founded in 1996 by German media company, Kirch to provide playout, multiplexing, satellite uplinks and other broadcast services to Kirch's Premiere pay-TV platform (now Sky Deutschland) and other private and public German broadcasters. In 2005, SES Astra (a subsidiary of SES Global, now SES) bought 100% of DPC from Premiere and the company renamed ASTRA Platform Services GmbH (APS). In 2012, to reflect the company's expanding worldwide reach, the name was changed to SES Platform Services. In February 2016, it was announced that SES Platform Services had agreed, subject to regulatory approvals, to purchase RR Media. The acquisition was completed in July 2016, with the merged company renamed MX1 and headed by Avi Cohen, the former CEO of RR Media. In October 2017, Cohen was replaced as CEO by Wilfred Urner, the former CEO of SES Platform Services, CEO of SES subsidiary, HD+ and Head of Media Platforms and Product Development, SES Video.