Since the Egyptian Revolution in 2011 and the Tunisian Revolution, social media, especially Facebook, Twitter, and YouTube, began to gain traction as a political tool in Africa. Various political actors have used social media to pursue a wide range of political objectives. State actors can use social media to encourage political discourse, campaign, or implement censorship and surveillance. Non-state actors, such as civil society organizations and opposition movements, can use social media to address political concerns and to organize widespread uprisings, such as the 2014 Burkinabé uprising. Meanwhile, extremist organizations can use social media to further their propaganda and recruitment. However, social media has been criticized for its limited accessibility and for facilitating the spread of misinformation, causing some skepticism about its effectiveness. Due to low entry barriers and user-generated content, social media provides a platform where people from different social classes can engage and interact with one another. Under traditional media, the public had limited opportunities to voice their political opinions. Social media enables people to both create and consume content. The public has become increasingly comfortable and confident in expressing political opinions online, often away from government scrutiny. Scholars argue that social media use has democratizing effects in African countries. == State actors == === Promoting political discourse === Through social media, the government and its citizens can discuss policy ideas, policy implementation, and political actions. Regardless of geographical location and distance, people are able to voice their opinions to the government. Social media includes citizens who were previously not able to express their discontent or share their ideas to the government. As state actors keep the public informed, social media can increase civic engagement. With more civic engagement, policies can be discussed without politicization. Before the commonplace use of social media, African countries faced weak feedback mechanisms that effectively excluded the average African citizen from policy discourse. In South Africa, the government uses social media to connect with constituencies. The South African president runs an official Twitter, Facebook, YouTube, and Flickr accounts to engage with the public. === Campaigning === Political parties also use social media for political campaigns during election periods. In South Africa, the ANC (African National Congress) and DA (Democratic Alliance) use social media for political purposes. These parties specifically use Facebook as a tool for campaigning and engaging with the public to improve their relationship with citizens. Nigerian President Goodluck Jonathan employed social media to campaign for the presidential election in 2011, which he won. When President Goodluck Jonathan announced his bid for the presidency on social media in 2010, it reached about 217,000 people. As his campaign progressed, President Goodluck Jonathan was able to increase his followers to half a million by early 2011. === Censorship & Surveillance === While state actors can use social media to encourage their party or discourse, social media can be used to censor and surveil citizens. For example, the ANC and DA use Facebook to monitor South Africans. The government is able to track down people who have spoken against the government and translate this information into physical action to stop any possibility of a revolution. Social media platforms can be shut down to manipulate the flow of information. In Chad, citizens cannot access information through online platforms. This censorship blocked "Facebook, Twitter, WhatsApp and Viber". In the Democratic Republic of Congo, the government shut down the internet before contested elections. In Zimbabwe, the government shut down the internet to hide civilian protests against fuel price increases. == Non-state actors == === Civil society organizations (CSOs) === Civil society organizations have also used social media networks in an effort to recruit supporters and communicate with the public. CSOs can use social media to mobilize people to support their cause, such as the Ghanaian Committee for Joint Action (CJA). In 2005 and 2006, the CJA gathered support to protest against the 50% fuel price increase. CSOs can play the role of a counterforce against state actors and state propaganda during times of crises, such as protests and military clashes. In some cases, CSOs release their own videos and photos on social media which challenges traditional forms of media. CSOs have also served to monitor elections to reduce corruption and violence during election day. For instance, the Zambian Bantu Watch started the #bantuwatch social media campaign to monitor the 2011 presidential election. Zambians used Facebook and Twitter to report polling station results to mitigate election fraud and election violence. In South Africa, CSOs created 'amandla.mobi' to campaign for public policies by creating petitions. Through 'amandla.mobi', CSOs are able to circulate petitions on social media to collect signatures. South African CSOs reported how social media helped their organizations to gain support and share ideas. However, CSOs struggle to attract media attention and often have to pay for media coverage. === Opposition forces against the government === Social media is also used by the public or opposition forces against the government. Through horizontal social media, organizing can lead to street protests and revolutions, some of which are successful. For instance, during the Egyptian revolution of 2011, "The Day of the Revolution Against Torture, Poverty, Corruption, and Unemployment" and "We Are All Khaled Said" gathered support against President Hosni Mubarak. In particular, "We Are All Khaled Said" had Egyptian citizens gather around the death of Khaled Said who was brutally tortured and killed by the Egyptian government because Said wanted to uncover government corruption. As unrest erupted into public demonstrations, President Hosni Mubarak was forced to resign. Witnessing the success of social media during the Egyptian revolution, the Tunisian Revolution, or the Jasmine Revolution, mobilized through Facebook and Twitter. Likewise, in South Africa, Malawi, and Mozambique, these countries have used social media as "new protest drums." Due to social media's low entry barrier, opposition forces against the government can facilitate political discourse that can lead to accountability. Whistleblowers and opposition forces are able to expose corruption through social media, where they face less repression while reaching a larger audience. For example, the youth of Zimbabwe and South Africa use Facebook to discuss politics without judgment. Specifically, in Zimbabwe, political youth used Facebook to avoid state surveillance. Social media is used as a supplemental tool for activism. In 2015, South African student activists started the hashtag #RhodesMustFall to push the issue of colonialism and racism at the forefront of the public. === Extremist organizations === Social media is easily accessible and created by user-based content. Therefore, marginalized groups are able to use social media to spread extremist ideas. For instance, Boko Haram created the Media Office of West Africa Province and perpetuated propaganda through Twitter and YouTube. Boko Haram's online propaganda campaign targets and persuades young dissuaded Nigerians to join their cause. It is important to note that social media has also been used against Boko Haram. In April 2014, Boko Haram kidnapped 276 schoolgirls and an international campaign fought for their return through #BringBackOurGirls. Another extremist group, Al-Shabaab, has created an online presence through Twitter and YouTube. Through these social media networks, Al-Shabaab recruits new members to their extremist group through their propaganda which emphasizes the group's successes. Albeit their efforts, Al-Shabaab has not been very successful in coordinating their members but they are successful in financing their group. Furthermore, the Islamic State of Iraq and the Levant (ISIL) use social media to target and recruit individuals to their cause. ISIL's social media usage is more diverse compared to Boko Haram and Al-Shabaab; ISIL uses "Facebook, Twitter, YouTube, WhatsApp, Telegram, JustPaste.it, Kik and Ask.fm." Since ISIL's Twitter accounts kept getting shut down, ISIL uses Telegram and WhatsApp chat rooms to privately conduct meetings. Due to the spread of extremist ideology, Zhuravskaya et al. acknowledge social media's potential to be misused. == Challenges == Although social media can be used as a political tool, it faces challenges in Africa. Due to low literacy rates in Africa, social media networks exclude many of the population members. In addition, lack of access to electricity and the internet can fur
Supersampling
Supersampling or supersampling anti-aliasing (SSAA) is a spatial anti-aliasing method, i.e. a method used to remove aliasing (jagged and pixelated edges, colloquially known as "jaggies") from images rendered in computer games or other computer programs that generate imagery. Aliasing occurs because unlike real-world objects, which have continuous smooth curves and lines, a computer screen shows the viewer a large number of small squares. These pixels all have the same size, and each one has a single color. A line can only be shown as a collection of pixels, and therefore appears jagged unless it is perfectly horizontal or vertical. The aim of supersampling is to reduce this effect. Color samples are taken at several instances inside the pixel (not just at the center as normal)—hence the term "supersampling"—and an average color value is calculated. This can for example be achieved by rendering the image at a much higher resolution than the one being displayed, then shrinking it to the desired size, using the extra pixels for calculation, with the result being a downsampled image with smoother transitions from one line of pixels to another along the edges of objects, but each pixel could also be supersampled using other strategies (see the Supersampling patterns section). The number of samples determines the quality of the output. == Motivation == Aliasing is manifested in the case of 2D images as moiré pattern and pixelated edges, colloquially known as "jaggies". Common signal processing and image processing knowledge suggests that to achieve perfect elimination of aliasing, proper spatial sampling at the Nyquist rate (or higher) after applying a 2D Anti-aliasing filter is required. As this approach would require a forward and inverse fourier transformation, computationally less demanding approximations like supersampling were developed to avoid domain switches by staying in the spatial domain ("image domain"). == Method == === Computational cost and adaptive supersampling === Supersampling is computationally expensive because it requires much greater video card memory and memory bandwidth, since the amount of buffer used is several times larger. A way around this problem is to use a technique known as adaptive supersampling, where only pixels at the edges of objects are supersampled. Initially only a few samples are taken within each pixel. If these values are very similar, only these samples are used to determine the color. If not, more are used. The result of this method is that a higher number of samples are calculated only where necessary, thus improving performance. === Supersampling patterns === When taking samples within a pixel, the sample positions have to be determined in some way. Although the number of ways in which this can be done is infinite, there are a few ways which are commonly used. ==== Grid ==== The simplest algorithm. The pixel is split into several sub-pixels, and a sample is taken from the center of each. It is fast and easy to implement. Although, due to the regular nature of sampling, aliasing can still occur if a low number of sub-pixels is used. ==== Random ==== Also known as stochastic sampling, it avoids the regularity of grid supersampling. However, due to the irregularity of the pattern, samples end up being unnecessary in some areas of the pixel and lacking in others. ==== Poisson disk ==== The Poisson disk sampling algorithm places the samples randomly, but then checks that any two are not too close. The end result is an even but random distribution of samples. The naive "dart throwing" algorithm is extremely slow for large data sets, which once limited its applications for real-time rendering. However, many fast algorithms now exist to generate Poisson disk noise, even those with variable density. The Delone set provides a mathematical description of such sampling. ==== Jittered ==== A modification of the grid algorithm to approximate the Poisson disk. A pixel is split into several sub-pixels, but a sample is not taken from the center of each, but from a random point within the sub-pixel. Congregation can still occur, but to a lesser degree. ==== Rotated grid ==== A 2×2 grid layout is used but the sample pattern is rotated to avoid samples aligning on the horizontal or vertical axis, greatly improving antialiasing quality for the most commonly encountered cases. For an optimal pattern, the rotation angle is arctan (1/2) (about 26.6°) and the square is stretched by a factor of √5/2, making it also a 4-queens solution.
Hanna Hajishirzi
Hannaneh Hajishirzi is an Iranian-American computer scientist specializing in natural language processing. She is Torode Family Professor in Computer Science & Engineering in the Paul G. Allen School of Computer Science and Engineering at the University of Washington, head of the H2Lab in the Allen School, and a senior director of natural language processing in the Allen Institute for AI. == Education and career == After a bachelor's degree from the Sharif University of Technology, Hajishirzi completed her Ph.D. in computer science in 2011, at the University of Illinois Urbana-Champaign. Her dissertation, Action-Centered Reasoning for Probabilistic Dynamic Systems, was supervised by Eyal Amir. After postdoctoral research at Disney Research in Pittsburgh, Hajishirzi joined the University of Washington in 2012, as a research scientist in electrical engineering. In 2015 she became a research assistant professor in electrical engineering. She obtained a regular-rank assistant professorship in 2018, at the same time becoming an AI Fellow in the Allen Institute for AI, where she became a senior director of research in 2021. She was promoted to associate professor in 2022 and to full professor in 2025. == Recognition == Hajishirzi was named as a Fellow of the Association for Computational Linguistics in 2025, "for significant contributions to question answering, scientific applications, multimodal artificial intelligence, and fully open language models". == Personal life == Hajishirzi is married to Ali Farhadi, the CEO of the Allen Institute for AI.
Margaret Mitchell (scientist)
Margaret Mitchell is a computer scientist who works on algorithmic bias and fairness in machine learning. She is most well known for her work on automatically removing undesired biases concerning demographic groups from machine learning models, as well as more transparent reporting of their intended use. == Education == Mitchell obtained a bachelor's degree in linguistics from Reed College, Portland, Oregon, in 2005. After having worked as a research assistant at the OGI School of Science and Engineering for two years, she subsequently obtained a Master's in Computational Linguistics from the University of Washington in 2009. She enrolled in a PhD program at the University of Aberdeen, where she wrote a doctoral thesis on the topic of Generating Reference to Visible Objects, graduating in 2013. == Career and research == Mitchell is best known for her work on fairness in machine learning and methods for mitigating algorithmic bias. This includes her work on introducing the concept of 'Model Cards' for more transparent model reporting, and methods for debiasing machine learning models using adversarial learning. Margaret Mitchell created the framework for recognizing and avoiding biases by testing with a variable for the group of interest, predictor and an adversary. In 2012, Mitchell joined the Human Language Technology Center of Excellence at Johns Hopkins University as a postdoctoral researcher, before taking up a position at Microsoft Research in 2013. At Microsoft, Mitchell was the research lead of the Seeing AI project, an app that offers support for the visually impaired by narrating texts and images. In November 2016, she became a senior research scientist at Google Research and Machine intelligence. While at Google, she founded and co-led the Ethical Artificial Intelligence team together with Timnit Gebru. In May 2018, she represented Google in the Partnership on AI. In February 2018, she gave a TED talk on "How we can build AI to help humans, not hurt us". In January 2021, after Timnit Gebru's termination from Google, Mitchell reportedly used a script to search through her corporate account and download emails that allegedly documented discriminatory incidents involving Gebru. An automated system locked Mitchell's account in response. In response to media attention Google claimed that she "exfiltrated thousands of files and shared them with multiple external accounts". After a five-week investigation, Mitchell was fired. Prior to her dismissal, Mitchell had been a vocal advocate for diversity at Google, and had voiced concerns about research censorship at the company. In late 2021, she joined AI start-up Hugging Face. Mitchell is a co-founder of Widening NLP, a special interest group within the Association for Computational Linguistics (ACL) seeking to increase the proportion of women and minorities working in natural language processing; and Computational Linguistics and Clinical Psychology, an annual workshop within the ACL that brings together clinicians and computational linguists to advance the state of the art in clinical psychology.
Supervised learning
In machine learning, supervised learning (SL) is a type of machine learning paradigm where an algorithm learns to map input data to a specific output based on example input-output pairs. This process involves training a statistical model using labeled data, meaning each piece of input data is provided with the correct output. The term "supervised" refers to the role of a teacher or supervisor who provides this training data, guiding the algorithm towards correct predictions. For instance, if you want a model to identify cats in images, supervised learning would involve feeding it many images of cats (inputs) that are explicitly labeled "cat" (outputs). The goal of supervised learning is for the trained model to accurately predict the output for new, unseen data. This requires the algorithm to effectively generalize from the training examples, a quality measured by its generalization error. Supervised learning is commonly used for tasks like classification (predicting a category, e.g., spam or not spam) and regression (predicting a continuous value, e.g., house prices). == Steps to follow == To solve a given problem of supervised learning, the following steps must be performed: Determine the type of training samples. Before doing anything else, the user should decide what kind of data is to be used as a training set. In the case of handwriting analysis, for example, this might be a single handwritten character, an entire handwritten word, an entire sentence of handwriting, or a full paragraph of handwriting. Gather a training set. The training set needs to be representative of the real-world use of the function. Thus, a set of input objects is gathered together with corresponding outputs, either from human experts or from measurements. Determine the input feature representation of the learned function. The accuracy of the learned function depends strongly on how the input object is represented. Typically, the input object is transformed into a feature vector, which contains a number of features that are descriptive of the object. The number of features should not be too large, because of the curse of dimensionality; but should contain enough information to accurately predict the output. Determine the structure of the learned function and corresponding learning algorithm. For example, one may choose to use support-vector machines or decision trees. Complete the design. Run the learning algorithm on the gathered training set. Some supervised learning algorithms require the user to determine certain control parameters. These parameters may be adjusted by optimizing performance on a subset (called a validation set) of the training set, or via cross-validation. Evaluate the accuracy of the learned function. After parameter adjustment and learning, the performance of the resulting function should be measured on a test set that is separate from the training set. == Algorithm choice == A wide range of supervised learning algorithms are available, each with its strengths and weaknesses. There is no single learning algorithm that works best on all supervised learning problems (see the No free lunch theorem). There are four major issues to consider in supervised learning: === Bias–variance tradeoff === A first issue is the tradeoff between bias and variance. Imagine that we have available several different, but equally good, training data sets. A learning algorithm is biased for a particular input x {\displaystyle x} if, when trained on each of these data sets, it is systematically incorrect when predicting the correct output for x {\displaystyle x} . A learning algorithm has high variance for a particular input x {\displaystyle x} if it predicts different output values when trained on different training sets. The prediction error of a learned classifier is related to the sum of the bias and the variance of the learning algorithm. Generally, there is a tradeoff between bias and variance. A learning algorithm with low bias must be "flexible" so that it can fit the data well. But if the learning algorithm is too flexible, it will fit each training data set differently, and hence have high variance. A key aspect of many supervised learning methods is that they are able to adjust this tradeoff between bias and variance (either automatically or by providing a bias/variance parameter that the user can adjust). === Function complexity and amount of training data === The second issue is of the amount of training data available relative to the complexity of the "true" function (classifier or regression function). If the true function is simple, then an "inflexible" learning algorithm with high bias and low variance will be able to learn it from a small amount of data. But if the true function is highly complex (e.g., because it involves complex interactions among many different input features and behaves differently in different parts of the input space), then the function will only be able to learn with a large amount of training data paired with a "flexible" learning algorithm with low bias and high variance. === Dimensionality of the input space === A third issue is the dimensionality of the input space. If the input feature vectors have large dimensions, learning the function can be difficult even if the true function only depends on a small number of those features. This is because the many "extra" dimensions can confuse the learning algorithm and cause it to have high variance. Hence, input data of large dimensions typically requires tuning the classifier to have low variance and high bias. In practice, if the engineer can manually remove irrelevant features from the input data, it will likely improve the accuracy of the learned function. In addition, there are many algorithms for feature selection that seek to identify the relevant features and discard the irrelevant ones. This is an instance of the more general strategy of dimensionality reduction, which seeks to map the input data into a lower-dimensional space prior to running the supervised learning algorithm. === Noise in the output values === A fourth issue is the degree of noise in the desired output values (the supervisory target variables). If the desired output values are often incorrect (because of human error or sensor errors), then the learning algorithm should not attempt to find a function that exactly matches the training examples. Attempting to fit the data too carefully leads to overfitting. You can overfit even when there are no measurement errors (stochastic noise) if the function you are trying to learn is too complex for your learning model. In such a situation, the part of the target function that cannot be modeled "corrupts" your training data – this phenomenon has been called deterministic noise. When either type of noise is present, it is better to go with a higher bias, lower variance estimator. In practice, there are several approaches to alleviate noise in the output values such as early stopping to prevent overfitting as well as detecting and removing the noisy training examples prior to training the supervised learning algorithm. There are several algorithms that identify noisy training examples and removing the suspected noisy training examples prior to training has decreased generalization error with statistical significance. === Other factors to consider === Other factors to consider when choosing and applying a learning algorithm include the following: Heterogeneity of the data. If the feature vectors include features of many different kinds (discrete, discrete ordered, counts, continuous values), some algorithms are easier to apply than others. Many algorithms, including support-vector machines, linear regression, logistic regression, neural networks, and nearest neighbor methods, require that the input features be numerical and scaled to similar ranges (e.g., to the [-1,1] interval). Methods that employ a distance function, such as nearest neighbor methods and support-vector machines with Gaussian kernels, are particularly sensitive to this. An advantage of decision trees is that they easily handle heterogeneous data. Redundancy in the data. If the input features contain redundant information (e.g., highly correlated features), some learning algorithms (e.g., linear regression, logistic regression, and distance-based methods) will perform poorly because of numerical instabilities. These problems can often be solved by imposing some form of regularization. Presence of interactions and non-linearities. If each of the features makes an independent contribution to the output, then algorithms based on linear functions (e.g., linear regression, logistic regression, support-vector machines, naive Bayes) and distance functions (e.g., nearest neighbor methods, support-vector machines with Gaussian kernels) generally perform well. However, if there are complex interactions among features, then algorithms such as decision trees and neural networks work better, becaus
CMU Pronouncing Dictionary
The CMU Pronouncing Dictionary (also known as CMUdict) is an open-source pronouncing dictionary originally created by the Speech Group at Carnegie Mellon University (CMU) for use in speech recognition research. CMUdict provides a mapping orthographic/phonetic for English words in their North American pronunciations. It is commonly used to generate representations for speech recognition (ASR), e.g. the CMU Sphinx system, and speech synthesis (TTS), e.g. the Festival system. CMUdict can be used as a training corpus for building statistical grapheme-to-phoneme (g2p) models that will generate pronunciations for words not yet included in the dictionary. The most recent release is 0.7b; it contains over 134,000 entries. An interactive lookup version is available. == Database format == The database is distributed as a plain text file with one entry to a line in the format "WORD
Deep Learning Studio
Deep Learning Studio is a software tool that aims to simplify the creation of deep learning models used in artificial intelligence. It is compatible with a number of open-source programming frameworks popularly used in artificial neural networks, including MXNet and Google's TensorFlow. Prior to the release of Deep Learning Studio in January 2017, proficiency in Python, among other programming languages, was essential in developing effective deep learning models. Deep Learning Studio sought to simplify the model creation process through a visual, drag-and-drop interface and the application of pre-trained learning models on available data. Irving, Texas–based Deep Cognition Inc. is the developer behind Deep Learning Studio. In 2017, the software allowed Deep Cognition to become a finalist for Best Innovation in Deep Learning in the Alconics Awards, which are given annually to the best artificial intelligence software. Deep Cognition launched version 2.0 of Deep Learning Studio at NVIDIA's GTC 2018 Conference in San Jose, California. Fremont, California–based computing products supplier Exxact Corp provides desktop computers specifically built to handle Deep Learning Studio workloads. == Features == Source: Deep Learning Studio is available in two versions: Desktop and Cloud, both of which are free software. The Desktop version is available on Windows and Ubuntu. The Cloud version is available in single-user and multi-user configurations. A Deep Cognition account is needed to access the Cloud version. Account registration is free. Deep Learning Studio can import existing Keras models; it also takes a data set as an input. Deep Learning Studio's AutoML feature allows automatic generation of deep learning models. More advanced users may choose to generate their own models using various types of layers and neural networks. Deep Learning Studio also has a library of loss functions and optimizers for use in hyperparameter tuning, a traditionally complicated area in neural network programming. Generated models can be trained using either CPUs or GPUs. Trained models can then be used for predictive analytics.