PyTorch is an open-source deep learning library, originally developed by Meta Platforms and currently developed with support from the Linux Foundation. The successor to Torch, PyTorch provides a high-level API that builds upon optimised, low-level implementations of deep learning algorithms and architectures, such as the Transformer, or SGD. Notably, this API simplifies model training and inference to a few lines of code. PyTorch allows for automatic parallelization of training and, internally, implements CUDA bindings that speed training further by leveraging GPU resources. PyTorch utilises the tensor as a fundamental data type, similarly to NumPy. Training is facilitated by a reversed automatic differentiation system, Autograd, that constructs a directed acyclic graph of the operations (and their arguments) executed by a model during its forward pass. With a loss, backpropagation is then undertaken. As of 2025, PyTorch remains one of the most popular deep learning libraries, alongside others such as TensorFlow and Keras. It can be installed using Anaconda package managers. A number of commercial deep learning architectures are built on top of PyTorch, including ChatGPT, Tesla Autopilot, Uber's Pyro, and Hugging Face's Transformers. == History == In 2001, Torch was written and released under a GPL. It was a machine-learning library written in C++ and CUDA, supporting methods including neural networks, support vector machines (SVM), hidden Markov models, etc. Around 2010, it was rewritten by Ronan Collobert, Clement Farabet and Koray Kavuckuoglu. This was known as Torch7 or LuaTorch. This was written so that the backend was in C and the frontend was in Lua. In mid-2016, some developers refactored it to decouple the frontend and the backend, with strong influence from torch-autograd and Chainer. In turn, torch-autograd was influenced by HIPS/autograd. Development on Torch7 ceased in 2018 and was subsumed by the PyTorch project. Meta (formerly known as Facebook) operates both PyTorch and Convolutional Architecture for Fast Feature Embedding (Caffe2), but models defined by the two frameworks were mutually incompatible. The Open Neural Network Exchange (ONNX) project was created by Meta and Microsoft in September 2017 to decouple deep learning frameworks from hardware-specific runtimes, allowing models to be converted between frameworks and optimized for execution providers like NVIDIA’s TensorRT. Caffe2 was merged into PyTorch at the end of March 2018. In September 2022, Meta announced that PyTorch would be governed by the independent PyTorch Foundation, a newly created subsidiary of the Linux Foundation. PyTorch 2.0 was released on 15 March 2023, introducing TorchDynamo, a Python-level compiler that makes code run up to two times faster, along with significant improvements in training and inference performance across major cloud platforms. == PyTorch tensors == PyTorch defines a class called Tensor (torch.Tensor) to store and operate on homogeneous multidimensional rectangular arrays of numbers. PyTorch supports various sub-types of multi-dimensional arrays, or Tensors. PyTorch Tensors are similar to NumPy Arrays, but can also be operated on by a CUDA-capable NVIDIA GPU. PyTorch has also been developing support for other GPU platforms, for example, AMD's ROCm and Apple's Metal Framework. == PyTorch neural networks == PyTorch defines a module called nn (torch.nn) to describe neural networks and to support training. This module offers a comprehensive collection of building blocks for neural networks, including various layers and activation functions, enabling the construction of complex models. Networks are built by inheriting from the torch.nn module and defining the sequence of operations in the forward() function. == PyTorch Serialized File Format == Pytorch can save and load models using its own file format, which is a ZIP64 archive containing the model weights in a Python pickle file, and other information such as the byte order. The file extensions .pt and .pth are commonly used for these files. == Example == The following program shows the low-level functionality of the library with a simple example. The following code block defines a neural network with linear layers using the nn module.
Error level analysis
Error level analysis (ELA) is the analysis of compression artifacts in digital data with lossy compression such as JPEG. == Principles == When used, lossy compression is normally applied uniformly to a set of data, such as an image, resulting in a uniform level of compression artifacts. Alternatively, the data may consist of parts with different levels of compression artifacts. This difference may arise from the different parts having been repeatedly subjected to the same lossy compression a different number of times, or the different parts having been subjected to different kinds of lossy compression. A difference in the level of compression artifacts in different parts of the data may therefore indicate that the data has been edited. In the case of JPEG, even a composite with parts subjected to matching compressions will have a difference in the compression artifacts. In order to make the typically faint compression artifacts more readily visible, the data to be analyzed is subjected to an additional round of lossy compression, this time at a known, uniform level, and the result is subtracted from the original data under investigation. The resulting difference image is then inspected manually for any variation in the level of compression artifacts. In 2007, N. Krawetz denoted this method "error level analysis". Additionally, digital data formats such as JPEG sometimes include metadata describing the specific lossy compression used. If in such data the observed compression artifacts differ from those expected from the given metadata description, then the metadata may not describe the actual compressed data, and thus indicate that the data have been edited. == Limitations == By its nature, data without lossy compression, such as a PNG image, cannot be subjected to error level analysis. Consequently, since editing could have been performed on data without lossy compression with lossy compression applied uniformly to the edited, composite data, the presence of a uniform level of compression artifacts does not rule out editing of the data. Additionally, any non-uniform compression artifacts in a composite may be removed by subjecting the composite to repeated, uniform lossy compression. Also, if the image color space is reduced to 256 colors or less, for example, by conversion to GIF, then error level analysis will generate useless results. More significant, the actual interpretation of the level of compression artifacts in a given segment of the data is subjective, and the determination of whether editing has occurred is therefore not robust. == Controversy == In May 2013, Dr Neal Krawetz used error level analysis on the 2012 World Press Photo of the Year and concluded on his Hacker Factor blog that it was "a composite" with modifications that "fail to adhere to the acceptable journalism standards used by Reuters, Associated Press, Getty Images, National Press Photographer's Association, and other media outlets". The World Press Photo organizers responded by letting two independent experts analyze the image files of the winning photographer and subsequently confirmed the integrity of the files. One of the experts, Hany Farid, said about error level analysis that "It incorrectly labels altered images as original and incorrectly labels original images as altered with the same likelihood". Krawetz responded by clarifying that "It is up to the user to interpret the results. Any errors in identification rest solely on the viewer". In May 2015, the citizen journalism team Bellingcat wrote that error level analysis revealed that the Russian Ministry of Defense had edited satellite images related to the Malaysia Airlines Flight 17 disaster. In a reaction to this, image forensics expert Jens Kriese said about error level analysis: "The method is subjective and not based entirely on science", and that it is "a method used by hobbyists". On his Hacker Factor Blog, the inventor of error level analysis Neal Krawetz criticized both Bellingcat's use of error level analysis as "misinterpreting the results" but also on several points Jens Kriese's "ignorance" regarding error level analysis.
Technical data management system
A technical data management system (TDMS) is a document management system (DMS) pertaining to the management of technical and engineering drawings and documents. Often the data are contained in 'records' of various forms, such as on paper, microfilms or digital media. Hence technical data management is also concerned with record management involving technical data. Technical document management systems are used within large organisations with large scale projects involving engineering. For example, a TDMS can be used for integrated steel plants (ISP), automobile factories, aero-space facilities, infrastructure companies, city corporations, research organisations, etc. In such organisations, technical archives or technical documentation centres are created as central facilities for effective management of technical data and records. TDMS functions are similar to that of conventional archive functions in concepts, except that the archived materials in this case are essentially engineering drawings, survey maps, technical specifications, plant and equipment data sheets, feasibility reports, project reports, operation and maintenance manuals, standards, etc. Document registration, indexing, repository management, reprography, etc. are parts of TDMS. Various kinds of sophisticated technologies such as document scanners, microfilming and digitization camera units, wide format printers, digital plotters, software, etc. are available, making TDMS functions an easier process than previous times. == Constituents of a technical data management system == Technical data refers to both scientific and technical information recorded and presented in any form or manner (excluding financial and management information). A Technical Data Management System is created within an organisation for archiving and sharing information such as technical specifications, datasheets and drawings. Similar to other types of data management system, a Technical Data Management System consists of the 4 crucial constituents mentioned below. === Data planning === Data plans (long-term or short-term) are constructed as the first essential step of a proper and complete TDMS. It is created to ultimately help with the 3 other constituents, data acquisition, data management and data sharing. A proper data plan should not exceed 2 pages and should address the following basics: Types of data (samples, experiment results, reports, drawings, etc.) and metadata (data that summarizes and describes other data. In this case, it refers to details such as sample sizes, experiment conditions and procedures, dates of reports, explanations of drawings, etc.) Means of researches and collections of data (field works, experiments in production lines, etc.) Costs of researches Policies for access, sharing (re-use within the organisation and re-distribution to the public) Proposals for archiving data and maintaining access to it === Data acquisition === Raw data is collected from primary sites of the organisations through the use of modern technologies. Please reference the table below for examples. The data collected is then transferred to technical data centres for data management. === Data management === After data acquisition, data is sorted out, whilst useful data is archived, unwanted data is disposed. When managing and archiving data, the features below of the data are considered. Names, labels, values and descriptions for variables and records. (In the case of TDMS, one example is names of equipments on an equipment datasheet) Derived data from the original data, with code, algorithm or command file used to create them. (In the case of TDMS, one example is an expectation report derived from the analysis of an equipment datasheet) Metadata associates with the data being archived === Data sharing === Archived and managed data are accessible to rightful entities. A proper and complete TDMS should share data to a suitable extent, under suitable security, in order to achieve optimal usage of data within the organisation. It aims for easy access when reused by other researchers and hence it enhances other research processes. Data is often referred in other tests and technical specifications, where new analysis is generated, managed and archived again. As a result, data is flowing within the organisation under effective management through the use of TDMS. == Advantages and disadvantages of usage of technical data management systems == There are strengths and weakness when using technical data management systems (TDMS) to archive data. Some of the advantages and disadvantages are listed below. === Advantages === ==== 1. Faster and easier data management ==== Since TDMS is integrated into the organisation's systems, whenever workers develop data files (SolidWorks, AutoCAD, Microsoft Word, etc.), they can also archive and manage data, linking what they need to their current work, at the same time they can also update the archives with useful data. This speeds up working processes and makes them more efficient. ==== 2. Increased security ==== All data files are centralized, hence internal and external data leakages are less likely to happen, and the data flow is more closely monitored. As a result, data in the organisation is more secured. ==== 3. Increased collaboration within the organisation ==== Since the data files are centralized and the data flow within the organisation increases, researchers and workers within the organisation are able to work on joint projects. More complex tasks can be performed for higher yields. ==== 4. Compatible to various formats of data ==== TDMS is compatible to many formats of data, from basic data like Microsoft Words to complex data like voice data. This enhances the quality of the management of data archived. === Disadvantages === ==== 1. Higher financial costs ==== Implementing TDMS into the organisation's systems involves monetary costs. Maintenance costs certain amount of human resources and money as well. These resources involve opportunity costs as they can be utilized in other aspects. ==== 2. Lower stability ==== Since TDMS manages and centralizes all the data the organisation processes, it links the working processes within the whole organisation together. It also increases the vulnerability of the organisation data network. If TDMS is not stable enough or when it is exposed to hacker and virus attacks, the organisation's data flow might shut down completely, affecting the work in an organisation-wide scale and leading to a lower stability as results. == Comparison between traditional data management approaches and technical data management systems == Test engineers and researchers are facing great challenges in turning complex test results and simulation data into usable information for higher yields of firms. These challenges are listed below. Increase in complication of designs Reduced in time and budgets available Higher quality is demanded === Traditional data management approaches === Many organisations are still applying the conventional file management systems, due to the difficulty in building a proper and complete archives for data management. The first approach is the simple file-folder system. This costs the problem of ineffectiveness as workers and researchers have to manually go through numerous layers of systems and files for the target data. Moreover, the target data may contain files with different formats and these files may not be stored in the same machine. These files are also easily lost if renamed or moved to another location. The second approach is conventional databases such as Oracle. These databases are capable of enabling easy search and access of data. However, a great drawback is that huge effort for preparing and modeling the data is required. For large-scale projects, huge monetary costs are induced, and extra IT human resources must be employed for constant handling, expanding and maintaining the inflexible system, which is custom for specific tasks, instead of all tasks. In the long-term, it is not cost-effective. === Technical data management systems (TDMS) === TDMS is developed based on 3 principles, flexible and organized file storage, self-scaling hybrid data index, and an interactive post-processing environment. The system in practical, mainly consists of 3 components, data files with essential and relevant Metadata, data finders for organizing and managing data regardless of files formats, and, a software of searching, analyzing and reporting. With metadata attached to original data files, the data finder can identify different related data files during searches, even if they are in different file formats. TDMS hence allows researchers to search for data like browsing the Internet. Last but not least, it can adapt to changes and update itself according to the changes, unlike databases. == Comparison between strong information systems and weak information systems == Complex organizations may need large amounts
Bibliographic database
A bibliographic database is a database of bibliographic records. This is an organised online collection of references to published written works like journal and newspaper articles, conference proceedings, reports, government and legal publications, patents and books. In contrast to library catalogue entries, a majority of the records in bibliographic databases describe articles and conference papers rather than complete monographs, and they generally contain very rich subject descriptions in the form of keywords, subject classification terms, or abstracts. A bibliographic database may cover a wide range of topics or one academic field like computer science. A significant number of bibliographic databases are marketed under a trade name by licensing agreement from vendors, or directly from their makers: the indexing and abstracting services. Many bibliographic databases have evolved into digital libraries, providing the full text of the organised contents:for instance CORE also organises and mirrors scholarly articles and OurResearch develops a search engine for open access content in Unpaywall. Others merge with non-bibliographic and scholarly databases to create more complete disciplinary search engine systems, such as Chemical Abstracts or Entrez. == History == Prior to the mid-20th century, individuals searching for published literature had to rely on printed bibliographic indexes, generated manually from index cards. During the early 1960s computers were used to digitize text for the first time; the purpose was to reduce the cost and time required to publish two American abstracting journals, the Index Medicus of the National Library of Medicine and the Scientific and Technical Aerospace Reports of the National Aeronautics and Space Administration (NASA). By the late 1960s, such bodies of digitized alphanumeric information, known as bibliographic and numeric databases, constituted a new type of information resource. Online interactive retrieval became commercially viable in the early 1970s over private telecommunications networks. The first services offered a few databases of indexes and abstracts of scholarly literature. These databases contained bibliographic descriptions of journal articles that were searchable by keywords in author and title, and sometimes by journal name or subject heading. The user interfaces were crude, the access was expensive, and searching was done by librarians on behalf of "end users".
Xulvi-Brunet–Sokolov algorithm
Xulvi-Brunet and Sokolov's algorithm generates networks with chosen degree correlations. This method is based on link rewiring, in which the desired degree is governed by parameter ρ. By varying this single parameter it is possible to generate networks from random (when ρ = 0) to perfectly assortative or disassortative (when ρ = 1). This algorithm allows to keep network's degree distribution unchanged when changing the value of ρ. == Assortative model == In assortative networks, well-connected nodes are likely to be connected to other highly connected nodes. Social networks are examples of assortative networks. This means that an assortative network has the property that almost all nodes with the same degree are linked only between themselves. The Xulvi-Brunet–Sokolov algorithm for this type of networks is the following. In a given network, two links connecting four different nodes are chosen randomly. These nodes are ordered by their degrees. Then, with probability ρ, the links are randomly rewired in such a way that one link connects the two nodes with the smaller degrees and the other connects the two nodes with the larger degrees. If one or both of these links already existed in the network, the step is discarded and is repeated again. Thus, there will be no self-connected nodes or multiple links connecting the same two nodes. Different degrees of assortativity of a network can be achieved by changing the parameter ρ. Assortative networks are characterized by highly connected groups of nodes with similar degree. As assortativity grows, the average path length and clustering coefficient increase. == Disassortative model == In disassortative networks, highly connected nodes tend to connect to less-well-connected nodes with larger probability than in uncorrelated networks. Examples of such networks include biological networks. The Xulvi-Brunet and Sokolov's algorithm for this type of networks is similar to the one for assortative networks with one minor change. As before, two links of four nodes are randomly chosen and the nodes are ordered with respect to their degrees. However, in this case, the links are rewired (with probability p) such that one link connects the highest connected node with the node with the lowest degree and the other link connects the two remaining nodes randomly with probability 1 − ρ. Similarly, if the new links already existed, the previous step is repeated. This algorithm does not change the degree of nodes and thus the degree distribution of the network.
Neural field
In machine learning, a neural field (also known as implicit neural representation, neural implicit, or coordinate-based neural network), is a mathematical field that is fully or partially parametrized by a neural network. Initially developed to tackle visual computing tasks, such as rendering or reconstruction (e.g., neural radiance fields), neural fields emerged as a promising strategy to deal with a wider range of problems, including surrogate modelling of partial differential equations, such as in physics-informed neural networks. Differently from traditional machine learning algorithms, such as feed-forward neural networks, convolutional neural networks, or transformers, neural fields do not work with discrete data (e.g. sequences, images, tokens), but map continuous inputs (e.g., spatial coordinates, time) to continuous outputs (i.e., scalars, vectors, etc.). This makes neural fields not only discretization independent, but also easily differentiable. Moreover, dealing with continuous data allows for a significant reduction in space complexity, which translates to a much more lightweight network. == Formulation and training == According to the universal approximation theorem, provided adequate learning, sufficient number of hidden units, and the presence of a deterministic relationship between the input and the output, a neural network can approximate any function to any degree of accuracy. Hence, in mathematical terms, given a field y = Φ ( x ) {\textstyle {\boldsymbol {y}}=\Phi ({\boldsymbol {x}})} , with x ∈ R n {\displaystyle {\boldsymbol {x}}\in \mathbb {R} ^{n}} and y ∈ R m {\displaystyle {\boldsymbol {y}}\in \mathbb {R} ^{m}} , a neural field Ψ θ {\displaystyle \Psi _{\theta }} , with parameters θ {\displaystyle {\boldsymbol {\theta }}} , is such that: Ψ θ ( x ) = y ^ ≈ y {\displaystyle \Psi _{\theta }({\boldsymbol {x}})={\hat {\boldsymbol {y}}}\approx {\boldsymbol {y}}} === Training === For supervised tasks, given N {\displaystyle N} examples in the training dataset (i.e., ( x i , y i ) ∈ D t r a i n , i = 1 , … , N {\displaystyle ({\boldsymbol {x_{i}}},{\boldsymbol {y_{i}}})\in {\mathcal {D_{train}}},i=1,\dots ,N} ), the neural field parameters can be learned by minimizing a loss function L {\displaystyle {\mathcal {L}}} (e.g., mean squared error). The parameters θ ~ {\displaystyle {\tilde {\theta }}} that satisfy the optimization problem are found as: θ ~ = argmin θ 1 N ∑ ( x i , y i ) ∈ D t r a i n L ( Ψ θ ( x i ) , y i ) {\displaystyle {\tilde {\boldsymbol {\theta }}}={\underset {\boldsymbol {\theta }}{\text{argmin}}}\;{\frac {1}{N}}\sum _{({\boldsymbol {x_{i}}},{\boldsymbol {y_{i}}})\in {\mathcal {D_{train}}}}{\mathcal {L}}(\Psi _{\theta }({\boldsymbol {x}}_{i}),{\boldsymbol {y}}_{i})} Notably, it is not necessary to know the analytical expression of Φ {\displaystyle \Phi } , for the previously reported training procedure only requires input-output pairs. Indeed, a neural field is able to offer a continuous and differentiable surrogate of the true field, even from purely experimental data. Moreover, neural fields can be used in unsupervised settings, with training objectives that depend on the specific task. For example, physics-informed neural networks may be trained on just the residual. === Spectral bias === As for any artificial neural network, neural fields may be characterized by a spectral bias (i.e., the tendency to preferably learn the low frequency content of a field), possibly leading to a poor representation of the ground truth. In order to overcome this limitation, several strategies have been developed. For example, SIREN uses sinusoidal activations, while the Fourier-features approach embeds the input through sines and cosines. == Conditional neural fields == In many real-world cases, however, learning a single field is not enough. For example, when reconstructing 3D vehicle shapes from Lidar data, it is desirable to have a machine learning model that can work with arbitrary shapes (e.g., a car, a bicycle, a truck, etc.). The solution is to include additional parameters, the latent variables (or latent code) z ∈ R d {\displaystyle {\boldsymbol {z}}\in \mathbb {R} ^{d}} , to vary the field and adapt it to diverse tasks. === Latent code production === When dealing with conditional neural fields, the first design choice is represented by the way in which the latent code is produced. Specifically, two main strategies can be identified: Encoder: the latent code is the output of a second neural network, acting as an encoder. During training, the loss function is the objective used to learn the parameters of both the neural field and the encoder. Auto-decoding: each training example has its own latent code, jointly trained with the neural field parameters. When the model has to process new examples (i.e., not originally present in the training dataset), a small optimization problem is solved, keeping the network parameters fixed and only learning the new latent variables. Since the latter strategy requires additional optimization steps at inference time, it sacrifices speed, but keeps the overall model smaller. Moreover, despite being simpler to implement, an encoder may harm the generalization capabilities of the model. For example, when dealing with a physical scalar field f : R 2 → R {\displaystyle f:\mathbb {R} ^{2}\rightarrow \mathbb {R} } (e.g., the pressure of a 2D fluid), an auto-decoder-based conditional neural field can map a single point to the corresponding value of the field, following a learned latent code z {\displaystyle {\boldsymbol {z}}} . However, if the latent variables were produced by an encoder, it would require access to the entire set of points and corresponding values (e.g. as a regular grid or a mesh graph), leading to a less robust model. === Global and local conditioning === In a neural field with global conditioning, the latent code does not depend on the input and, hence, it offers a global representation (e.g., the overall shape of a vehicle). However, depending on the task, it may be more useful to divide the domain of x {\displaystyle {\boldsymbol {x}}} in several subdomains, and learn different latent codes for each of them (e.g., splitting a large and complex scene in sub-scenes for a more efficient rendering). This is called local conditioning. === Conditioning strategies === There are several strategies to include the conditioning information in the neural field. In the general mathematical framework, conditioning the neural field with the latent variables is equivalent to mapping them to a subset θ ∗ {\displaystyle {\boldsymbol {\theta }}^{}} of the neural field parameters: θ ∗ = Γ ( z ) {\displaystyle {\boldsymbol {\theta }}^{}=\Gamma ({\boldsymbol {z}})} In practice, notable strategies are: Concatenation: the neural field receives, as input, the concatenation of the original input x {\displaystyle {\boldsymbol {x}}} with the latent codes z {\displaystyle {\boldsymbol {z}}} . For feed-forward neural networks, this is equivalent to setting θ ∗ {\displaystyle {\boldsymbol {\theta }}^{}} as the bias of the first layer and Γ ( z ) {\displaystyle \Gamma ({\boldsymbol {z}})} as an affine transformation. Hypernetworks: a hypernetwork is a neural network that outputs the parameters of another neural network. Specifically, it consists of approximating Γ ( z ) {\displaystyle \Gamma ({\boldsymbol {z}})} with a neural network Γ ^ γ ( z ) {\displaystyle {\hat {\Gamma }}_{\gamma }({\boldsymbol {z}})} , where γ {\displaystyle {\boldsymbol {\gamma }}} are the trainable parameters of the hypernetwork. This approach is the most general, as it allows to learn the optimal mapping from latent codes to neural field parameters. However, hypernetworks are associated to larger computational and memory complexity, due to the large number of trainable parameters. Hence, leaner approaches have been developed. For example, in the Feature-wise Linear Modulation (FiLM), the hypernetwork only produces scale and bias coefficients for the neural field layers. === Meta-learning === Instead of relying on the latent code to adapt the neural field to a specific task, it is also possible to exploit gradient-based meta-learning. In this case, the neural field is seen as the specialization of an underlying meta-neural-field, whose parameters are modified to fit the specific task, through a few steps of gradient descent. An extension of this meta-learning framework is the CAVIA algorithm, that splits the trainable parameters in context-specific and shared groups, improving parallelization and interpretability, while reducing meta-overfitting. This strategy is similar to the auto-decoding conditional neural field, but the training procedure is substantially different. == Applications == Thanks to the possibility of efficiently modelling diverse mathematical fields with neural networks, neural fields have been applied to a wide range of problems: 3D scene reconstruction: neural fields can be used to model t
Sparse identification of non-linear dynamics
Sparse identification of nonlinear dynamics (SINDy) is a data-driven algorithm for obtaining dynamical systems from data. Given a series of snapshots of a dynamical system and its corresponding time derivatives, SINDy performs a sparsity-promoting regression (such as LASSO and sparse Bayesian inference) on a library of nonlinear candidate functions of the snapshots against the derivatives to find the governing equations. This procedure relies on the assumption that most physical systems only have a few dominant terms which dictate the dynamics, given an appropriately selected coordinate system and quality training data. It has been applied to identify the dynamics of fluids, based on proper orthogonal decomposition, as well as other complex dynamical systems, such as biological networks. == Mathematical Overview == First, consider a dynamical system of the form x ˙ = d d t x ( t ) = f ( x ( t ) ) , {\displaystyle {\dot {\textbf {x}}}={\frac {d}{dt}}{\textbf {x}}(t)={\textbf {f}}({\textbf {x}}(t)),} where x ( t ) ∈ R n {\displaystyle {\textbf {x}}(t)\in \mathbb {R} ^{n}} is a state vector (snapshot) of the system at time t {\displaystyle t} and the function f ( x ( t ) ) {\displaystyle {\textbf {f}}({\textbf {x}}(t))} defines the equations of motion and constraints of the system. The time derivative may be either prescribed or numerically approximated from the snapshots. With x {\displaystyle {\textbf {x}}} and x ˙ {\displaystyle {\dot {\textbf {x}}}} sampled at m {\displaystyle m} equidistant points in time ( t 1 , t 2 , ⋯ , t m {\displaystyle t_{1},t_{2},\cdots ,t_{m}} ), these can be arranged into matrices of the form X = [ x T ( t 1 ) x T ( t 2 ) ⋮ x T ( t m ) ] = [ x 1 ( t 1 ) x 2 ( t 1 ) ⋯ x n ( t 1 ) x 1 ( t 2 ) x 2 ( t 2 ) ⋯ x n ( t 2 ) ⋮ ⋮ ⋱ ⋮ x 1 ( t m ) x 2 ( t m ) ⋯ x n ( t m ) ] , {\displaystyle {\bf {{X}={\begin{bmatrix}\mathbf {x} ^{\mathsf {T}}(t_{1})\\\mathbf {x} ^{\mathsf {T}}(t_{2})\\\vdots \\\mathbf {x} ^{\mathsf {T}}(t_{m})\end{bmatrix}}={\begin{bmatrix}x_{1}(t_{1})&x_{2}(t_{1})&\cdots &x_{n}(t_{1})\\x_{1}(t_{2})&x_{2}(t_{2})&\cdots &x_{n}(t_{2})\\\vdots &\vdots &\ddots &\vdots \\x_{1}(t_{m})&x_{2}(t_{m})&\cdots &x_{n}(t_{m})\end{bmatrix}},}}} and similarly for X ˙ {\displaystyle {\dot {\mathbf {X} }}} . Next, a library Θ ( X ) {\displaystyle \mathbf {\Theta } (\mathbf {X} )} of nonlinear candidate functions of the columns of X {\displaystyle {\textbf {X}}} is constructed, which may be constant, polynomial, or more exotic functions (like trigonometric and rational terms, and so on): Θ ( X ) = [ | | | | | | 1 X X 2 X 3 ⋯ sin ( X ) cos ( X ) ⋯ | | | | | | ] {\displaystyle \ \ \ {\bf {{\Theta }({\bf {{X})={\begin{bmatrix}\vline &\vline &\vline &\vline &&\vline &\vline &\\1&{\bf {X}}&{\bf {{X}^{2}}}&{\bf {{X}^{3}}}&\cdots &\sin({\bf {{X})}}&\cos({\bf {{X})}}&\cdots \\\vline &\vline &\vline &\vline &&\vline &\vline &\end{bmatrix}}}}}}} The number of possible model structures from this library is combinatorially high. f ( x ( t ) ) {\displaystyle {\textbf {f}}({\textbf {x}}(t))} is then substituted by Θ ( X ) {\displaystyle {\bf {{\Theta }({\textbf {X}})}}} and a vector of coefficients Ξ = [ ξ 1 ξ 2 ⋯ ξ n ] {\displaystyle {\bf {{\Xi }=\left[{\bf {{\xi }_{1}{\bf {{\xi }_{2}\cdots {\bf {{\xi }_{n}}}}}}}\right]}}} determining the active terms in f ( x ( t ) ) {\displaystyle {\textbf {f}}({\textbf {x}}(t))} : X ˙ = Θ ( X ) Ξ {\displaystyle {\dot {\bf {X}}}={\bf {{\Theta }({\bf {{X}){\bf {\Xi }}}}}}} Because only a few terms are expected to be active at each point in time, an assumption is made that f ( x ( t ) ) {\displaystyle {\textbf {f}}({\textbf {x}}(t))} admits a sparse representation in Θ ( X ) {\displaystyle {\bf {{\Theta }({\textbf {X}})}}} . This then becomes an optimization problem in finding a sparse Ξ {\displaystyle {\bf {\Xi }}} which optimally embeds X ˙ {\displaystyle {\dot {\textbf {X}}}} . In other words, a parsimonious model is obtained by performing least squares regression on the system (4) with sparsity-promoting ( L 1 {\displaystyle L_{1}} ) regularization ξ k = arg min ξ k ′ | | X ˙ k − Θ ( X ) ξ k ′ | | 2 + λ | | ξ k ′ | | 1 , {\displaystyle {\bf {{\xi }_{k}={\underset {\bf {{\xi }'_{k}}}{\arg \min }}\left|\left|{\dot {\bf {X}}}_{k}-{\bf {{\Theta }({\bf {{X}){\bf {{\xi }'_{k}}}}}}}\right|\right|_{2}+\lambda \left|\left|{\bf {{\xi }'_{k}}}\right|\right|_{1},}}} where λ {\displaystyle \lambda } is a regularization parameter. Finally, the sparse set of ξ k {\displaystyle {\bf {{\xi }_{k}}}} can be used to reconstruct the dynamical system: x ˙ k = Θ ( x ) ξ k {\displaystyle {\dot {x}}_{k}={\bf {{\Theta }({\bf {{x}){\bf {{\xi }_{k}}}}}}}}