Generative literature is poetry or fiction that is automatically generated, often using computers. It is a genre of electronic literature, and also related to generative art. John Clark's Latin Verse Machine (1830–1843) is probably the first example of mechanised generative literature, while Christopher Strachey's love letter generator (1952) is the first digital example. With the large language models (LLMs) of the 2020s, generative literature is becoming increasingly common. == Definitions == Hannes Bajohr defines generative literature as literature involving "the automatic production of text according to predetermined parameters, usually following a combinatory, sometimes aleatory logic, and it emphasizes the production rather than the reception of the work (unlike, say, hypertext)." In his book Electronic Literature, Scott Rettberg connects generative literature to avant-garde literary movements like Dada, Surrealism, Oulipo and Fluxus. Bajohr argues that conceptual art is also an important reference. == Paradigms of generative literature == Bajohr describes two main paradigms of generative literature: the sequential paradigm, where the text generation is "executed as a sequence of rule-steps" and employs linear algorithms, and the connectionist paradigm, which is based on neural nets. The latter leads to what Bajohr calls a algorithmic empathy: "a non-anthropocentric empathy aimed not at the psychological states of the artists but at understanding the process of the work’s material production." == Poetry generation == The first examples of automated generative literature are poetry: John Clark's mechanical Latin Verse Machine (1830–1843) produced lines of hexameter verse in Latin, and Christopher Strachey's love letter generator (1952), programmed on the Manchester Mark 1 computer, generated short, satirical love letters. Examples of generative poetry using artificial neural networks include David Jhave Johnston's ReRites. == Narrative generation == Story generators have often followed specific narratological theories of how stories are constructed. An early example is Grimes' Fairy Tales, the "first to take a grammar-based approach and the first to operationalize Propp's famous model." Mike Sharples and Rafael Peréz y Peréz's book Story Machines gives a detailed history of story generation. Storyland by Nanette Wylde is an example of generative narrative. Jonathan Baillehache compares Storyland to Surrealist writing. Baillehache states, "When compared to earlier uses of chance operation in literature, a piece like this one resembles some of the automatic writings produced by André Breton and Philippe Soupault in their collective work The Magnetic Fields. . . The difference between Nanette Wylde’s Storyland and Breton and Soupault’s Magnetic Fields is that the former is produced according to a computational algorithm involving randomizers and user interaction, and the latter by two free-wheeling human subjects."
DataScene
DataScene is a scientific graphing, animation, data analysis, and real-time data monitoring software package. It was developed with the Common Language Infrastructure technology and the GDI+ graphics library. With the two Common Language Runtime engines - the .Net and Mono frameworks - DataScene runs on all major operating systems. With DataScene, the user can plot 39 types 2D & 3D graphs (e.g., Area graph, Bar graph, Boxplot graph, Pie graph, Line graph, Histogram graph, Surface graph, Polar graph, Water Fall graph, etc.), manipulate, print, and export graphs to various formats (e.g., Bitmap, WMF/EMF, JPEG, PNG, GIF, TIFF, PostScript, and PDF), analyze data with different mathematical methods (fitting curves, calculating statics, FFT, etc.), create chart animations for presentations (e.g. with PowerPoint), classes, and web pages, and monitor and chart real-time data. == History == DataScene was first released (version 1.0) in March 2009 for the Windows platform and the .Net 2.0 framework. Since version 2.0, DataScene has been ported to the Mono framework 2.6 and all Linux and Unix/X11 operating systems. Cyberwit offers free licensing for the Express edition of DataScene.
Win–stay, lose–switch
In psychology, game theory, statistics, and machine learning, win–stay, lose–switch (also win–stay, lose–shift or Pavlov, named after Ivan Pavlov) is a heuristic learning strategy used to model learning in decision situations. It was first invented as an improvement over randomization in bandit problems. It was later applied to the prisoner's dilemma in order to model the evolution of altruism. In most versions, it starts either with a cooperate, then proceeds as always, or starts with a "probe" of cooperate-defect-cooperate to determine the other player's strategy. A mutual cooperation is regarded as a win. The learning rule bases its decision only on the outcome of the previous play. Outcomes are divided into successes (wins) and failures (losses). If the play on the previous round resulted in a success, then the agent plays the same strategy on the next round. Alternatively, if the play resulted in a failure the agent switches to another action. A large-scale empirical study of players of the game rock, paper, scissors shows that a variation of this strategy is adopted by real-world players of the game, instead of the Nash equilibrium strategy of choosing entirely at random between the three options.
Relief (feature selection)
Relief is an algorithm developed by Kenji Kira and Larry Rendell in 1992 that takes a filter-method approach to feature selection that is notably sensitive to feature interactions. It was originally designed for application to binary classification problems with discrete or numerical features. Relief calculates a feature score for each feature which can then be applied to rank and select top scoring features for feature selection. Alternatively, these scores may be applied as feature weights to guide downstream modeling. Relief feature scoring is based on the identification of feature value differences between nearest neighbor instance pairs. If a feature value difference is observed in a neighboring instance pair with the same class (a 'hit'), the feature score decreases. Alternatively, if a feature value difference is observed in a neighboring instance pair with different class values (a 'miss'), the feature score increases. The original Relief algorithm has since inspired a family of Relief-based feature selection algorithms (RBAs), including the ReliefF algorithm. Beyond the original Relief algorithm, RBAs have been adapted to (1) perform more reliably in noisy problems, (2) generalize to multi-class problems (3) generalize to numerical outcome (i.e. regression) problems, and (4) to make them robust to incomplete (i.e. missing) data. To date, the development of RBA variants and extensions has focused on four areas; (1) improving performance of the 'core' Relief algorithm, i.e. examining strategies for neighbor selection and instance weighting, (2) improving scalability of the 'core' Relief algorithm to larger feature spaces through iterative approaches, (3) methods for flexibly adapting Relief to different data types, and (4) improving Relief run efficiency. Their strengths are that they are not dependent on heuristics, they run in low-order polynomial time, and they are noise-tolerant and robust to feature interactions, as well as being applicable for binary or continuous data; however, it does not discriminate between redundant features, and low numbers of training instances fool the algorithm. == Relief Algorithm == Take a data set with n instances of p features, belonging to two known classes. Within the data set, each feature should be scaled to the interval [0 1] (binary data should remain as 0 and 1). The algorithm will be repeated m times. Start with a p-long weight vector (W) of zeros. At each iteration, take the feature vector (X) belonging to one random instance, and the feature vectors of the instance closest to X (by Euclidean distance) from each class. The closest same-class instance is called 'near-hit', and the closest different-class instance is called 'near-miss'. Update the weight vector such that W i = W i − ( x i − n e a r H i t i ) 2 + ( x i − n e a r M i s s i ) 2 , {\displaystyle W_{i}=W_{i}-(x_{i}-\mathrm {nearHit} _{i})^{2}+(x_{i}-\mathrm {nearMiss} _{i})^{2},} where i {\displaystyle i} indexes the components and runs from 1 to p. Thus the weight of any given feature decreases if it differs from that feature in nearby instances of the same class more than nearby instances of the other class, and increases in the reverse case. After m iterations, divide each element of the weight vector by m. This becomes the relevance vector. Features are selected if their relevance is greater than a threshold τ. Kira and Rendell's experiments showed a clear contrast between relevant and irrelevant features, allowing τ to be determined by inspection. However, it can also be determined by Chebyshev's inequality for a given confidence level (α) that a τ of 1/sqrt(αm) is good enough to make the probability of a Type I error less than α, although it is stated that τ can be much smaller than that. Relief was also described as generalizable to multinomial classification by decomposition into a number of binary problems. == ReliefF Algorithm == Kononenko et al. propose a number of updates to Relief. Firstly, they find the near-hit and near-miss instances using the Manhattan (L1) norm rather than the Euclidean (L2) norm, although the rationale is not specified. Furthermore, they found taking the absolute differences between xi and near-hiti, and xi and near-missi to be sufficient when updating the weight vector (rather than the square of those differences). === Reliable probability estimation === Rather than repeating the algorithm m times, implement it exhaustively (i.e. n times, once for each instance) for relatively small n (up to one thousand). Furthermore, rather than finding the single nearest hit and single nearest miss, which may cause redundant and noisy attributes to affect the selection of the nearest neighbors, ReliefF searches for k nearest hits and misses and averages their contribution to the weights of each feature. k can be tuned for any individual problem. === Incomplete data === In ReliefF, the contribution of missing values to the feature weight is determined using the conditional probability that two values should be the same or different, approximated with relative frequencies from the data set. This can be calculated if one or both features are missing. === Multi-class problems === Rather than use Kira and Rendell's proposed decomposition of a multinomial classification into a number of binomial problems, ReliefF searches for k near misses from each different class and averages their contributions for updating W, weighted with the prior probability of each class. == Other Relief-based Algorithm Extensions/Derivatives == The following RBAs are arranged chronologically from oldest to most recent. They include methods for improving (1) the core Relief algorithm concept, (2) iterative approaches for scalability, (3) adaptations to different data types, (4) strategies for computational efficiency, or (5) some combination of these goals. For more on RBAs see these book chapters or this most recent review paper. === RRELIEFF === Robnik-Šikonja and Kononenko propose further updates to ReliefF, making it appropriate for regression. === Relieved-F === Introduced deterministic neighbor selection approach and a new approach for incomplete data handling. === Iterative Relief === Implemented method to address bias against non-monotonic features. Introduced the first iterative Relief approach. For the first time, neighbors were uniquely determined by a radius threshold and instances were weighted by their distance from the target instance. === I-RELIEF === Introduced sigmoidal weighting based on distance from target instance. All instance pairs (not just a defined subset of neighbors) contributed to score updates. Proposed an on-line learning variant of Relief. Extended the iterative Relief concept. Introduced local-learning updates between iterations for improved convergence. === TuRF (a.k.a. Tuned ReliefF) === Specifically sought to address noise in large feature spaces through the recursive elimination of features and the iterative application of ReliefF. === Evaporative Cooling ReliefF === Similarly seeking to address noise in large feature spaces. Utilized an iterative `evaporative' removal of lowest quality features using ReliefF scores in association with mutual information. === EReliefF (a.k.a. Extended ReliefF) === Addressing issues related to incomplete and multi-class data. === VLSReliefF (a.k.a. Very Large Scale ReliefF) === Dramatically improves the efficiency of detecting 2-way feature interactions in very large feature spaces by scoring random feature subsets rather than the entire feature space. === ReliefMSS === Introduced calculation of feature weights relative to average feature 'diff' between instance pairs. === SURF === SURF identifies nearest neighbors (both hits and misses) based on a distance threshold from the target instance defined by the average distance between all pairs of instances in the training data. Results suggest improved power to detect 2-way epistatic interactions over ReliefF. === SURF (a.k.a. SURFStar) === SURF extends the SURF algorithm to not only utilized 'near' neighbors in scoring updates, but 'far' instances as well, but employing inverted scoring updates for 'far instance pairs. Results suggest improved power to detect 2-way epistatic interactions over SURF, but an inability to detect simple main effects (i.e. univariate associations). === SWRF === SWRF extends the SURF algorithm adopting sigmoid weighting to take distance from the threshold into account. Also introduced a modular framework for further developing RBAs called MoRF. === MultiSURF (a.k.a. MultiSURFStar) === MultiSURF extends the SURF algorithm adapting the near/far neighborhood boundaries based on the average and standard deviation of distances from the target instance to all others. MultiSURF uses the standard deviation to define a dead-band zone where 'middle-distance' instances do not contribute to scoring. Evidence suggests MultiSURF performs best in detecting pure 2-way feature interactions. === Reli
Clustering illusion
The clustering illusion is the tendency to erroneously consider the inevitable "streaks" or "clusters" arising in small samples from random distributions to be non-random. The illusion is caused by a human tendency to underpredict the amount of variability likely to appear in a small sample of random or pseudorandom data. Thomas Gilovich, an early author on the subject, argued that the effect occurs for different types of random dispersions. Some might perceive patterns in stock market price fluctuations over time, or clusters in two-dimensional data such as the locations of impact of World War II V-1 flying bombs on maps of London. Although Londoners developed specific theories about the pattern of impacts within London, a statistical analysis by R. D. Clarke originally published in 1946 showed that the impacts of V-2 rockets on London were a close fit to a random distribution. == Similar biases == Using this cognitive bias in causal reasoning may result in the Texas sharpshooter fallacy, in which differences in data are ignored and similarities are overemphasized. More general forms of erroneous pattern recognition are pareidolia and apophenia. Related biases are the illusion of control which the clustering illusion could contribute to, and insensitivity to sample size in which people don't expect greater variation in smaller samples. A different cognitive bias involving misunderstanding of chance streams is the gambler's fallacy. == Possible causes == Daniel Kahneman and Amos Tversky explained this kind of misprediction as being caused by the representativeness heuristic (which itself they also first proposed).
Deep image compositing
Deep image compositing is a way of compositing and rendering digital images that emerged in the mid-2010s. In addition to the usual color and opacity channels a notion of spatial depth is created. This allows multiple samples in the depth of the image to make up the final resulting color. This technique produces high quality results and removes artifacts around edges that could not be dealt with otherwise. == Deep data == Deep data is encoded by advanced 3D renderers into an image that samples information about the path each rendered pixel takes along the z axis extending outward from the virtual camera through space, including the color and opacity of every non-opaque surface or volume it passes through along the way, as well as neighboring samples. It might be considered somewhat analogous to the way ray tracing generates simulated photon paths through such mediums; however, ray tracing and other traditional rendering techniques generally produce images that contain only three or four channels of color and opacity values per pixel, flattened into a two dimensional frame. Depth maps, on the other hand, contain z axis information encoded in a grayscale image. Each level of gray represents a different slice of the z space. The "thickness" of each slice is determined at time of render, allowing for more or less depth fidelity depending on how deep the scene is. Depth maps have been a boon to compositors for blending 3D renders with live action and practical elements. To be useful, the map must have high enough bit depth to encode separation between close-to-camera objects and objects near infinity. Most 3D software packages are now capable of generating 16-bit and 32-bit depth maps, providing up to 2 billion depth levels. Depth maps do not however include transparency information about non-opaque surfaces or volumes and as such, objects beyond and viewed through these semi- or fully-transparent objects will have no depth information of their own and may not get composited or blurred correctly. Even the popular addition of cryptomattes to many post-production and VFX studios' pipelines, while providing separate color-coded ID shapes for individual elements in a rendered scene to further bridge the gap between CGI and compositing, don't allow for the nearly automated and fully non-linear workflows that deep data does. This is because deep images encapsulate enough 3D information that normally time-intensive tasks such as rotoscoping with numerous holdout mattes for complex interactions between moving characters and semi-transparent environmental volumes like smoke or water, are essentially trivial. Instead of going through that process, multiple mattes could easily be generated from a single set of deep images with no need to re-render every matte element and background for each case. In addition to that efficiency and flexibility, deep data images inherently provide much higher visual quality in common areas that have been difficult with traditional renders, such as the motion-blurred edges of characters with semi-transparent elements like hair. One downside to the use of deep images is their substantial file size, since they encode a relatively enormous amount of data per frame compared to even multichannel formats such as OpenEXR. === Function-based (integrated) === The data is stored as a function of depth. This results in a function curve that can be used to look up the data at any arbitrary depth. Manipulating the data is harder. === Sample-based (deintegrated) === Each sample is considered as an independent piece and can so be manipulated easily. To make sure the data is representing the right detail, an additional expand value needs to be introduced. == Generating deep data == 3D renderers produce the necessary data as a part of the rendering pipeline. Samples are gathered in depth and then combined. The deep data can be written out before this happens and so is nothing new to the process. Generating deep data from camera data needs a proper depth map. This is used in a couple of cases but still not accurate enough for detailed representation. For basic holdout task this can be sufficient though. == Compositing deep data images == Deep images can be composited like regular images. The depth component makes it easier to determine the layering order. Traditionally this had to be input by the user. Deep images have that information for themselves and need no user input. Edge artifacts are reduced as transparent pixels have more data to work with. == History == Deep Images have been around in 3D rendering packages for quite a while now. The use of them for holdouts was first done at several VFX houses in shaders. Holdout mattes can be generated at render time. Using them in a more interactive manner was started recently by several companies, SideFX integrated it in their Houdini software and facilities like Industrial Light & Magic, DreamWorks Animation, Weta, AnimalLogic and DRD studios have implemented interactive solutions. In 2014 the Academy of Motion Picture Arts and Sciences honored the technology with its annual SciTech awards. Dr. Peter Hillman for the long-term development and continued advancement of innovative, robust and complete toolsets for deep compositing and to Colin Doncaster, Johannes Saam, Areito Echevarria, Janne Kontkanen and Chris Cooper for the development, prototyping and promotion of technologies and workflows for deep compositing. == Resources == Pixar Paper Deep Image Paper Video tutorial of Deep Imaging as used on 2012 film Rise of the Planet of the Apes, Nuke compositing software Deep Compositing Course Deep Image File Format at Google Code Academy Award for the Technology Theory of Deep Pixels OpenEXR Deep Pixels
Sharpness aware minimization
Sharpness Aware Minimization (SAM) is an optimization algorithm used in machine learning that aims to improve model generalization. The method seeks to find model parameters that are located in regions of the loss landscape with uniformly low loss values, rather than parameters that only achieve a minimal loss value at a single point. This approach is described as finding "flat" minima instead of "sharp" ones. The rationale is that models trained this way are less sensitive to variations between training and test data, which can lead to better performance on unseen data. The algorithm was introduced in a 2020 paper by a team of researchers including Pierre Foret, Ariel Kleiner, Hossein Mobahi, and Behnam Neyshabur. == Underlying Principle == SAM modifies the standard training objective by minimizing a "sharpness-aware" loss. This is formulated as a minimax problem where the inner objective seeks to find the highest loss value in the immediate neighborhood of the current model weights, and the outer objective minimizes this value: min w max ‖ ϵ ‖ p ≤ ρ L train ( w + ϵ ) + λ ‖ w ‖ 2 2 {\displaystyle \min _{w}\max _{\|\epsilon \|_{p}\leq \rho }L_{\text{train}}(w+\epsilon )+\lambda \|w\|_{2}^{2}} In this formulation: w {\displaystyle w} represents the model's parameters (weights). L train {\displaystyle L_{\text{train}}} is the loss calculated on the training data. ϵ {\displaystyle \epsilon } is a perturbation applied to the weights. ρ {\displaystyle \rho } is a hyperparameter that defines the radius of the neighborhood (an L p {\displaystyle L_{p}} ball) to search for the highest loss. An optional L2 regularization term, scaled by λ {\displaystyle \lambda } , can be included. A direct solution to the inner maximization problem is computationally expensive. SAM approximates it by taking a single gradient ascent step to find the perturbation ϵ {\displaystyle \epsilon } . This is calculated as: ϵ ( w ) = ρ ∇ L train ( w ) ‖ ∇ L train ( w ) ‖ 2 {\displaystyle \epsilon (w)=\rho {\frac {\nabla L_{\text{train}}(w)}{\|\nabla L_{\text{train}}(w)\|_{2}}}} The optimization process for each training step involves two stages. First, an "ascent step" computes a perturbed set of weights, w adv = w + ϵ ( w ) {\displaystyle w_{\text{adv}}=w+\epsilon (w)} , by moving towards the direction of the highest local loss. Second, a "descent step" updates the original weights w {\displaystyle w} using the gradient calculated at these perturbed weights, ∇ L train ( w adv ) {\displaystyle \nabla L_{\text{train}}(w_{\text{adv}})} . This update is typically performed using a standard optimizer like SGD or Adam. == Application and Performance == SAM has been applied in various machine learning contexts, primarily in computer vision. Research has shown it can improve generalization performance in models such as Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) on image datasets including ImageNet, CIFAR-10, and CIFAR-100. The algorithm has also been found to be effective in training models with noisy labels, where it performs comparably to methods designed specifically for this problem. Some studies indicate that SAM and its variants can improve out-of-distribution (OOD) generalization, which is a model's ability to perform well on data from distributions not seen during training. Other areas where it has been applied include gradual domain adaptation and mitigating overfitting in scenarios with repeated exposure to training examples. == Limitations == A primary limitation of SAM is its computational cost. By requiring two gradient computations (one for the ascent and one for the descent) per optimization step, it approximately doubles the training time compared to standard optimizers. The theoretical convergence properties of SAM are still under investigation. Some research suggests that with a constant step size, SAM may not converge to a stationary point. The accuracy of the single gradient step approximation for finding the worst-case perturbation may also decrease during the training process. The effectiveness of SAM can also be domain-dependent. While it has shown benefits for computer vision tasks, its impact on other areas, such as GPT-style language models where each training example is seen only once, has been reported as limited in some studies. Furthermore, while SAM seeks flat minima, some research suggests that not all flat minima necessarily lead to good generalization. The algorithm also introduces the neighborhood size ρ {\displaystyle \rho } as a new hyperparameter, which requires tuning. == Research, Variants, and Enhancements == Active research on SAM focuses on reducing its computational overhead and improving its performance. Several variants have been proposed to make the algorithm more efficient. These include methods that attempt to parallelize the two gradient computations, apply the perturbation to only a subset of parameters, or reduce the number of computation steps required. Other approaches use historical gradient information or apply SAM steps intermittently to lower the computational burden. To improve performance and robustness, variants have been developed that adapt the neighborhood size based on model parameter scales (Adaptive SAM or ASAM) or incorporate information about the curvature of the loss landscape (Curvature Regularized SAM or CR-SAM). Other research explores refining the perturbation step by focusing on specific components of the gradient or combining SAM with techniques like random smoothing. Theoretical work continues to analyze the algorithm's behavior, including its implicit bias towards flatter minima and the development of broader frameworks for sharpness-aware optimization that use different measures of sharpness.