Class activation mapping methods are explainable AI (XAI) techniques used to visualize the regions of an input image that are the most relevant for a particular task, especially image classification, in convolutional neural networks (CNNs). These methods generate heatmaps by weighting the feature maps from a convolutional layer according to their relevance to the target class. In the field of artificial intelligence, generically defined as "the effort to automate intellectual tasks normally performed by humans", machine learning and deep learning were created. They both use statistical and computational methods to learn patterns from data, reducing the need for manually coded rules. Machine learning models are trained on input data and the known respective answers, learning the underlying patterns or structures present in the data. Traditional Machine learning algorithms employ manually designed feature sets, posing a direct link between machine learning designers and employed features. Deep learning is a subfield of machine learning, based on the concept of successive layers of representation, in which the data is progressively unfolded in different ways, to extract relevant and informative patterns in data analysis. Deep learning algorithms are defined as feature learning algorithms automatically learning hierarchical feature representations from raw data, extracting increasingly abstract features through multiple layers. CNNs are a specific architecture of deep learning models, designed to process spatially structured data, such as images, exploiting a series of convolution, non-linear activation and pooling operations to extract relevant features, contained in the so-called feature maps from input data. CNNs have demonstrated to be highly effective in a variety of computer vision and image processing tasks. CNNs (and deep learning models more broadly) are described as black boxes due to their complex and non-transparent internal layers of representation. The need for clearer indications on its internal working and decision-making process gave birth to XAI techniques. Among the proposed XAI techniques for computer vision tasks, Class activation mapping methods can show which pixels in an input image are important to the predicted logit for a class of interest, in a classification task. Class activation mapping methods were originally developed for class-discriminative scenarios to visualize which parts of the input image influenced the classification decision, namely to visually highlight the regions of those feature maps that contribute most strongly to the prediction of a given class. More advanced versions of these methods are not limited to image classification tasks, but have been extended also to several vision-related tasks, such as object detection, image captioning, visual question answering and image segmentation. == Background == The following methods laid the groundwork for the class activation maps approaches, forming the conceptual basis of using gradients to highlight class-discriminative regions. === Class model visualization and saliency maps for convolutional neural networks === The class model visualization and image-specific saliency maps approaches have been presented in the foundational work "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps" by Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman and it generalizes the deconvnet method by Zeiler and Fergus. Class model visualization synthesizes an artificial input image that strongly activates the output neurons associated with a target class. Given a trained, fixed model, this method starts with a zero-initialized image, backpropagates the gradients from the class score to the image pixels, updates the image pixels increasing the specific class scores and it repeats the pixel updating process, showing an encoded (idealized version) prototype of the class of interest. Image-specific class saliency visualization method provides a visual explanation by highlighting the most relevant pixels in an image for predicting a certain class C of interest. This is done by computing the gradient of the class score with respect to the input image, I 0 , {\displaystyle I_{0},} w = ∂ S C ∂ I | I 0 {\displaystyle w=\left.{\frac {\partial S_{C}}{\partial I}}\right|_{I_{0}}} approximating the model locally (around I 0 {\displaystyle I_{0}} ) as linear, using a first-order Taylor expansion: S C ( I ) ≈ w C T I + b {\displaystyle S_{C}(I)\approx w_{C}^{T}I+b} . The magnitude of w C {\displaystyle w_{C}} , the gradient, indicates the importancy of the pixels: larger gradients suggest greater influence on the prediction. Once the gradient is known, the saliency map is defined as the maximum absolute gradient across the color channels: M i j = m a x C | ∂ S C ∂ I i j C | {\displaystyle M_{ij}=max_{C}\left|{\frac {\partial S_{C}}{\partial I_{ij}^{C}}}\right|} resulting in an saliency map (i.e. heatmap). === Guided backpropagation === The concept of guided backpropagation can be traced for the first time in the paper by Springenberg et al. "Striving For Simplicity: The All Convolutional Net" and also this method builds upon the work by Zeiler and Fergus "Visualizing and Understanding Convolutional Networks". Guided backpropagation core is to understand what a CNN is learning, by visualizing the patterns that activate more strongly individual neurons (or filters), in architectures which do not rely on max-pooling layer. When propagating gradients back through a rectified linear unit (ReLU), guided backpropagation passes the gradient if and only if the input to the ReLU was positive (forward pass) and the output gradient is positive (backward signal), tackling both inactive neurons, negative gradients and suppressing the noise. The result displays sharper, high-resolution visualizations of what each neuron is responding to. Guided backpropagation represents a simple and practical method for model interpretability, helping understand how and where neural networks detect semantic concepts across layers. Moreover, it can be applied to any network architecture, due to its working principle. == Base versions == Class activation mapping and gradient-weighted class activation mapping are the original and most widely used methods for visual explanations in convolutional neural networks. These methods serve as the foundation for many later developments in explainable AI. Notation: In this article, the symbols i and j represent integer indices that disappear inside sums or averages, while x and y are the continuous (or up-sampled integer) coordinates of the final heat-map that is plotted. === Class activation mapping (CAM) === Class activation mapping (CAM) was the first, and the original, version of CAM methods, and it gave the name to the whole category. The approach was firstly introduced by Zhou et al. in their seminal work "Learning Deep Features for Discriminative Localization". This approach achieves class-specific heatmaps by modifying image classification CNN architectures, replacing fully-connected layers with convolutional layers and a final global average pooling layer. Its main scope is to localize and highlight discriminative regions of an input image that a CNN uses to identify a particular class, without needing explicit bounding box annotations. ==== Global average pooling (GAP) ==== Global average pooling (GAP) represents the key element in the original CAM approach. It is a dimensionality reduction technique and, similarly to other pooling layers, it allows the downsampling of the feature maps, calculating representative values for a specific region of the feature map. The particularity of GAP is that it calculates a single value for an entire feature map, significantly reducing the model dimensions. ==== Mathematical description ==== The mathematical description considers as its key the combination of convolutional and GAP layers. In CAM, it is mandatory to have the GAP layer after the last convolutional layer and before the final linear classifier layer. This last element of the architecture connects the output logits (the network predictions) y C {\displaystyle y^{C}} , to the GAP values, with its respective fine-tuned weights, w k C {\displaystyle w_{k}^{C}} . Considering A k {\displaystyle A^{k}} as the last feature maps of the last convolutional layer, GAP produces one value for each feature map, by averaging all the matrix elements (i, j) of the feature map: F k = 1 m n ∑ i = 1 m ∑ j = 1 n A i j k {\displaystyle F^{k}={\frac {1}{mn}}\sum _{i=1}^{m}\sum _{j=1}^{n}A_{ij}^{k}} with A k = [ A 11 k A 12 k ⋯ A 1 n k A 21 k A 22 k ⋯ A 2 n k ⋮ ⋮ ⋱ ⋮ A m 1 k A m 2 k ⋯ A m n k ] = { A i j k ∣ 1 ≤ i ≤ m , 1 ≤ j ≤ n } {\displaystyle A^{k}={\begin{bmatrix}A_{11}^{k}&A_{12}^{k}&\cdots &A_{1n}^{k}\\A_{21}^{k}&A_{22}^{k}&\cdots &A_{2n}^{k}\\\vdots &\vdots &\ddots &\vdots \\A_{m1}^{k}&A_{m2}^{k}&\cdots &A_{mn}^{k}\end{bmatrix}}=\left\{A_{
Perusall
Perusall is a social web annotation tool intended for use by students at schools and universities. It allows users to annotate the margins of a text in a virtual group setting that is similar to social media—with upvoting, emojis, chat functionality, and notification. It also includes automatic AI grading. == History == Perusall began as a research project at Harvard University. It later became an educational product for students and teachers. As of 2024, Perusall states more than 5 million students have used the tool at over 5,000 educational institutions in 112 countries." == Functionality == Perusall integrates with learning management systems such as Moodle, Canvas and Blackboard to aid with collaborative annotation. The tool supports annotation of a range of media including text, images, equations, videos, PDFs and snapshots of webpages.
Chris Olah
Christopher Olah (born 1992 or 1993) is a Canadian machine learning researcher and a co-founder of Anthropic. He is known for his work on neural network interpretability, particularly mechanistic interpretability, and for research and tools that visualise internal representations in neural networks. In 2025, Forbes reported he had become a billionaire due to his ownership in Anthropic. == Early life and education == Olah was born in Canada. According to Wired, he left university at age 18 without earning a degree and later received a Thiel Fellowship, which supported him in pursuing independent work. == Career == Olah has worked on interpretability research at Google Brain, OpenAI, and Anthropic. Time called him one of the pioneers of mechanistic interpretability and noted that he pursued this research line first at Google, then at OpenAI, and later at Anthropic, which he co-founded. Wired reported that Olah was involved in neural network visualisation work including DeepDream in 2015, as part of efforts to better understand what neural networks learn. Later coverage linked him to more structured interpretability approaches such as "activation atlases". The Verge covered activation atlases as a collaboration between Google and OpenAI researchers to help inspect neural network representations. At Anthropic, Olah has been identified in major press coverage as leading interpretability work aimed at mapping internal "features" in large language models and relating interpretability findings to AI safety. Quanta Magazine has also quoted Olah in reporting on interpretability and the internal structure of modern language models. Time included Olah in its TIME100 AI list in 2024. === Vatican address on AI ethics === On May 25, 2026, Olah spoke at the Vatican during the official presentation of Magnifica Humanitas, the first encyclical of Pope Leo XIV, which addresses artificial intelligence and human dignity. Olah said AI could lead to large-scale displacement of human labor and exacerbate global inequality. He said the commercial and geopolitical incentives driving frontier AI labs often conflict with the public good, and described AI systems as "grown" rather than strictly engineered. Olah called for external moral oversight from religious institutions, scholars, and civil society to hold the technology sector accountable.
R2ML
The REWERSE Rule Markup Language (R2ML) is developed by the REWERSE Working Group I1 for the purpose of rules interchange between different systems and tools. == Scope == An XML based rule language; Support for: integrity rules, derivation rules, production rules and reaction rules; Integrate functional languages (such as OCL) with Datalog languages (such as SWRL); Serialization and interchange of rules by specific software tools; Integrating rule reasoning with actual server side technologies; Deploying, publishing and communicating rules in a network. == Design principles == Modeled using MDA; Rule concepts defined with the help of MOF/UML; Required to accommodate: Web naming concepts, such as URIs and XML namespaces; The ontological distinction between objects and data values; The datatype concepts of RDF and user-defined datatypes; Actions (following OMG PRR submission); Events; EBNF abstract syntax; XML based concrete syntax validated by an XML Schema; Allowing different semantics for rules.
Goal node (computer science)
In computer science, a goal node is a node in a graph that meets defined criteria for success or termination. Heuristical artificial intelligence algorithms, like A and B, attempt to reach such nodes in optimal time by defining the distance to the goal node. When the goal node is reached, A defines the distance to the goal node as 0 and all other nodes' distances as positive values.
Multimedia database
A Multimedia database (MMDB) is a collection of related for multimedia data. The multimedia data include one or more primary media data types such as text, images, graphic objects (including drawings, sketches and illustrations) animation sequences, audio and video. A Multimedia Database Management System (MMDBMS) is a framework that manages different types of data potentially represented in a wide diversity of formats on a wide array of media sources. It provides support for multimedia data types, and facilitate for creation, storage, access, query and control of a multimedia database. == Contents of MMDB == A Multimedia Database (MMDB) hosts one or more multimedia data types (i.e. text, images, graphic objects, audio, video, animation sequences). These data types are broadly categorized into three classes: Static media (time-independent: image and graphic object). Dynamic media (time-dependent: audio, video and animation). Dimensional media(3D game and computer aided drafting programs). === Comparison of multimedia data types === Additionally, a Multimedia Database (MMDB) needs to manage additional information pertaining to the actual multimedia data. The information is about the following: Media data: the actual data representing an object. Media format data: information about the format of the media data after it goes through the acquisition, processing, and encoding phases. Media keyword data: the keyword descriptions, usually relating to the generation of the media data. Media feature data: content dependent data such as contain information about the distribution of colours, the kinds of textures and the different shapes present in an image. The last three types are called metadata as they describe several different aspects of the media data. The media keyword data and media feature data are used as indices for searching purpose. The media format data is used to present the retrieved information. == Requirements of Multimedia databases == Like the traditional databases, Multimedia databases should address the following requirements: Integration Data items do not need to be duplicated for different programs invocations Data independence Separate the database and the management from the application programs Concurrency control Allows concurrent transactions Persistence Data objects can be saved and re-used by different transactions and program invocations Privacy Access and authorization control Integrity control Ensures database consistency between transactions Recovery Failures of transactions should not affect the persistent data storage Query support Allows easy querying of multimedia data Multimedia databases should have the ability to uniformly query data (media data, textual data) represented in different formats and have the ability to simultaneously query different media sources and conduct classical database operations across them. (Query support) They should have the ability to retrieve media objects from a local storage device in a good manner. (Storage support) They should have the ability to take the response generated by a query and develop a presentation of that response in terms of audio-visual media and have the ability to deliver this presentation. (Presentation and delivery support) == Issues and challenges == Multimedia data consists of a variety of media formats or file representations including TIFF, BMP, PPT, IVUE, FPX, JPEG, MPEG, AVI, MID, WAV, DOC, GIF, EPS, PNG, etc. Because of restrictions on the conversion from one format to the other, the use of the data in a specific format has been limited as well. Usually, the data size of multimedia is large such as video; therefore, multimedia data often require a large storage. Multimedia database consume a lot of processing time, as well as bandwidth. Some multimedia data types such as video, audio, and animation sequences have temporal requirements that have implications on their storage, manipulation and presentation, but images, video and graphics data have special constraints in terms of their content. == Application areas == Examples of multimedia database application areas: Digital Libraries News-on-Demand Video-on-Demand Music database Geographic Information Systems (GIS) Telemedicine
General Data Protection Regulation
The General Data Protection Regulation (Regulation (EU) 2016/679), abbreviated GDPR, is a European Union regulation on information privacy in the European Union (EU) and the European Economic Area (EEA). The GDPR is an important component of EU privacy law and human rights law, in particular Article 8(1) of the Charter of Fundamental Rights of the European Union. It also governs the transfer of personal data outside the EU and EEA. The GDPR's goals are to enhance individuals' control and rights over their personal information and to simplify the regulations for international business. It supersedes the Data Protection Directive 95/46/EC and, among other things, simplifies the terminology. The European Parliament and Council of the European Union adopted the GDPR on 14 April 2016, to become effective on 25 May 2018. As an EU regulation (instead of a directive), the GDPR has direct legal effect and does not require transposition into national law. However, it also provides flexibility for individual member states to modify (derogate from) some of its provisions. As an example of the Brussels effect, the regulation became a model for many other laws around the world, including in Brazil, Japan, Singapore, South Africa, South Korea, Sri Lanka, and Thailand. After leaving the European Union, the United Kingdom enacted its "UK GDPR", identical to the GDPR. The California Consumer Privacy Act (CCPA), adopted on 28 June 2018, has many similarities with the GDPR. == Contents == The GDPR 2016 has eleven chapters, concerning general provisions, principles, rights of the data subject, duties of data controllers or processors, transfers of personal data to third-party countries, supervisory authorities, cooperation among member states, remedies, liability or penalties for breach of rights, provisions related to specific processing situations, and miscellaneous final provisions. The GDPR also contains 173 recitals purposed to clarify scope and rationale for the regulatory provisions, as well as its legislative intents – Recital 4, for instance, begins by saying that the processing of personal data should be "designed to serve mankind". === General provisions === The regulation applies if the data controller, or processor, or the data subject (person) is based in the EU. The regulation also applies to organisations based outside the EU if they collect or process personal data of individuals located inside the EU. The regulation does not apply to the processing of data by private persons provided that the purpose has no connection to a professional or commercial activity." (Recital 18). According to the European Commission, "Personal data is information that relates to an identified or identifiable individual. If you cannot directly identify an individual from that information, then you need to consider whether the individual is still identifiable. You should take into account the information you are processing together with all the means reasonably likely to be used by either you or any other person to identify that individual." The precise definitions of terms such as "personal data", "processing", "data subject", "controller", and "processor" are stated in Article 4. The regulation does not purport to apply to the processing of personal data for national security activities or law enforcement of the EU; however, industry groups concerned about facing a potential conflict of laws have questioned whether Article 48 could be invoked to seek to prevent a data controller subject to a third country's laws from complying with a legal order from that country's law enforcement, judicial, or national security authorities to disclose to such authorities the personal data of an EU person, regardless of whether the data resides in or out of the EU. Article 48 states that any judgement of a court or tribunal and any decision of an administrative authority of a third country requiring a controller or processor to transfer or disclose personal data may not be recognised or enforceable in any manner unless based on an international agreement, like a mutual legal assistance treaty in force between the requesting third (non-EU) country and the EU or a member state. The data protection reform package also includes a separate Data Protection Directive for the police and criminal justice sector that provides rules on personal data exchanges at State level, Union level, and international levels. A single set of rules applies to all EU member states. Each member state establishes an independent supervisory authority (SA) to hear and investigate complaints, sanction administrative offences, etc. SAs in each member state co-operate with other SAs, providing mutual assistance and organising joint operations. If a business has multiple establishments in the EU, it must have a single SA as its "lead authority", based on the location of its "main establishment" where the main processing activities take place. The lead authority thus acts as a "one-stop shop" to supervise all the processing activities of that business throughout the EU. A European Data Protection Board (EDPB) co-ordinates the SAs. EDPB thus replaces the Article 29 Data Protection Working Party. There are exceptions for data processed in an employment context or in national security that still might be subject to individual country regulations. === Principles and lawful purposes === Article 5 sets out six principles relating to the lawfulness of processing personal data. The first of these specifies that data must be processed lawfully, fairly and in a transparent manner. Article 6 develops this principle by specifying that personal data may not be processed unless there is at least one legal basis for doing so. The other principles refer to "purpose limitation", "data minimisation", "accuracy", "storage limitation", and "integrity and confidentiality". Article 6 states that the lawful purposes are: (a) If the data subject has given consent to the processing of his or her personal data; (b) To fulfill contractual obligations with a data subject, or for tasks at the request of a data subject who is in the process of entering into a contract; (c) To comply with a data controller's legal obligations; (d) To protect the vital interests of a data subject or another individual; (e) To perform a task in the public interest or in official authority; (f) For the legitimate interests of a data controller or a third party, unless these interests are overridden by interests of the data subject or her or his rights according to the Charter of Fundamental Rights (especially in the case of children). If informed consent is used as the lawful basis for processing, consent must have been explicit for data collected and each purpose data is used for. Consent must be a specific, freely given, plainly worded, and unambiguous affirmation given by the data subject; an online form which has consent options structured as an opt-out selected by default is a violation of the GDPR, as the consent is not unambiguously affirmed by the user. In addition, multiple types of processing may not be "bundled" together into a single affirmation prompt, as this is not specific to each use of data, and the individual permissions are not freely given. (Recital 32). Data subjects must be allowed to withdraw this consent at any time, and the process of doing so must not be harder than it was to opt in. A data controller may not refuse service to users who decline consent to processing that is not strictly necessary in order to use the service. Consent for children, defined in the regulation as being less than 16 years old (although with the option for member states to individually make it as low as 13 years old), must be given by the child's parent or custodian, and verifiable. If consent to processing was already provided under the Data Protection Directive, a data controller does not have to re-obtain consent if the processing is documented and obtained in compliance with the GDPR's requirements (Recital 171). === Rights of the data subject === ==== Transparency and modalities ==== Article 12 requires the data controller to provide information to the "data subject in a concise, transparent, intelligible and easily accessible form, using clear and plain language, in particular for any information addressed specifically to a child." ==== Information and access ==== The right of access (Article 15) is a data subject right. It gives people the right to access their personal data and information about how this personal data is being processed. A data controller must provide, upon request, an overview of the categories of data that are being processed as well as a copy of the actual data; furthermore, the data controller has to inform the data subject on details about the processing, such as the purposes of the processing, with whom the data is shared, and how it acquired the data. A data subject must be able to transfer personal data from one electro