Proximal policy optimization

Proximal policy optimization (PPO) is a reinforcement learning (RL) algorithm for training an intelligent agent. Specifically, it is a policy gradient method, often used for deep RL when the policy network is very large. == History == The predecessor to PPO, Trust Region Policy Optimization (TRPO), was published in 2015. It addressed the instability issue of another algorithm, the Deep Q-Network (DQN), by using the trust region method to limit the KL divergence between the old and new policies. However, TRPO uses the Hessian matrix (a matrix of second derivatives) to enforce the trust region, but the Hessian is inefficient for large-scale problems. PPO was published in 2017. It was essentially an approximation of TRPO that does not require computing the Hessian. The KL divergence constraint was approximated by simply clipping the policy gradient. Since 2018, PPO was the default RL algorithm at OpenAI. PPO has been applied to many areas, such as controlling a robotic arm, beating professional players at Dota 2 (OpenAI Five), and playing Atari games. == TRPO == TRPO, the predecessor of PPO, is an on-policy algorithm. It can be used for environments with either discrete or continuous action spaces. The pseudocode is as follows: Input: initial policy parameters θ 0 {\textstyle \theta _{0}} , initial value function parameters ϕ 0 {\textstyle \phi _{0}} Hyperparameters: KL-divergence limit δ {\textstyle \delta } , backtracking coefficient α {\textstyle \alpha } , maximum number of backtracking steps K {\textstyle K} for k = 0 , 1 , 2 , … {\textstyle k=0,1,2,\ldots } do Collect set of trajectories D k = { τ i } {\textstyle {\mathcal {D}}_{k}=\left\{\tau _{i}\right\}} by running policy π k = π ( θ k ) {\textstyle \pi _{k}=\pi \left(\theta _{k}\right)} in the environment. Compute rewards-to-go R ^ t {\textstyle {\hat {R}}_{t}} . Compute advantage estimates, A ^ t {\textstyle {\hat {A}}_{t}} (using any method of advantage estimation) based on the current value function V ϕ k {\textstyle V_{\phi _{k}}} . Estimate policy gradient as g ^ k = 1 | D k | ∑ τ ∈ D k ∑ t = 0 T ∇ θ log ⁡ π θ ( a t ∣ s t ) | θ k A ^ t {\displaystyle {\hat {g}}_{k}=\left.{\frac {1}{\left|{\mathcal {D}}_{k}\right|}}\sum _{\tau \in {\mathcal {D}}_{k}}\sum _{t=0}^{T}\nabla _{\theta }\log \pi _{\theta }\left(a_{t}\mid s_{t}\right)\right|_{\theta _{k}}{\hat {A}}_{t}} Use the conjugate gradient algorithm to compute x ^ k ≈ H ^ k − 1 g ^ k {\displaystyle {\hat {x}}_{k}\approx {\hat {H}}_{k}^{-1}{\hat {g}}_{k}} where H ^ k {\textstyle {\hat {H}}_{k}} is the Hessian of the sample average KL-divergence. Update the policy by backtracking line search with θ k + 1 = θ k + α j 2 δ x ^ k T H ^ k x ^ k x ^ k {\displaystyle \theta _{k+1}=\theta _{k}+\alpha ^{j}{\sqrt {\frac {2\delta }{{\hat {x}}_{k}^{T}{\hat {H}}_{k}{\hat {x}}_{k}}}}{\hat {x}}_{k}} where j ∈ { 0 , 1 , 2 , … K } {\textstyle j\in \{0,1,2,\ldots K\}} is the smallest value which improves the sample loss and satisfies the sample KL-divergence constraint. Fit value function by regression on mean-squared error: ϕ k + 1 = arg ⁡ min ϕ 1 | D k | T ∑ τ ∈ D k ∑ t = 0 T ( V ϕ ( s t ) − R ^ t ) 2 {\displaystyle \phi _{k+1}=\arg \min _{\phi }{\frac {1}{\left|{\mathcal {D}}_{k}\right|T}}\sum _{\tau \in {\mathcal {D}}_{k}}\sum _{t=0}^{T}\left(V_{\phi }\left(s_{t}\right)-{\hat {R}}_{t}\right)^{2}} typically via some gradient descent algorithm. == PPO == The pseudocode is as follows: Input: initial policy parameters θ 0 {\textstyle \theta _{0}} , initial value function parameters ϕ 0 {\textstyle \phi _{0}} for k = 0 , 1 , 2 , … {\textstyle k=0,1,2,\ldots } do Collect set of trajectories D k = { τ i } {\textstyle {\mathcal {D}}_{k}=\left\{\tau _{i}\right\}} by running policy π k = π ( θ k ) {\textstyle \pi _{k}=\pi \left(\theta _{k}\right)} in the environment. Compute rewards-to-go R ^ t {\textstyle {\hat {R}}_{t}} . Compute advantage estimates, A ^ t {\textstyle {\hat {A}}_{t}} (using any method of advantage estimation) based on the current value function V ϕ k {\textstyle V_{\phi _{k}}} . Update the policy by maximizing the PPO-Clip objective: θ k + 1 = arg ⁡ max θ 1 | D k | T ∑ τ ∈ D k ∑ t = 0 T min ( π θ ( a t ∣ s t ) π θ k ( a t ∣ s t ) A π θ k ( s t , a t ) , g ( ϵ , A π θ k ( s t , a t ) ) ) {\displaystyle \theta _{k+1}=\arg \max _{\theta }{\frac {1}{\left|{\mathcal {D}}_{k}\right|T}}\sum _{\tau \in {\mathcal {D}}_{k}}\sum _{t=0}^{T}\min \left({\frac {\pi _{\theta }\left(a_{t}\mid s_{t}\right)}{\pi _{\theta _{k}}\left(a_{t}\mid s_{t}\right)}}A^{\pi _{\theta _{k}}}\left(s_{t},a_{t}\right),\quad g\left(\epsilon ,A^{\pi _{\theta _{k}}}\left(s_{t},a_{t}\right)\right)\right)} typically via stochastic gradient ascent with Adam. Fit value function by regression on mean-squared error: ϕ k + 1 = arg ⁡ min ϕ 1 | D k | T ∑ τ ∈ D k ∑ t = 0 T ( V ϕ ( s t ) − R ^ t ) 2 {\displaystyle \phi _{k+1}=\arg \min _{\phi }{\frac {1}{\left|{\mathcal {D}}_{k}\right|T}}\sum _{\tau \in {\mathcal {D}}_{k}}\sum _{t=0}^{T}\left(V_{\phi }\left(s_{t}\right)-{\hat {R}}_{t}\right)^{2}} typically via some gradient descent algorithm. Like all policy gradient methods, PPO is used for training an RL agent whose actions are determined by a differentiable policy function by gradient ascent. Intuitively, a policy gradient method takes small policy update steps, so the agent can reach higher and higher rewards in expectation. Policy gradient methods may be unstable: A step size that is too big may direct the policy in a suboptimal direction, thus having little possibility of recovery; a step size that is too small lowers the overall efficiency. To solve the instability, PPO implements a clip function that constrains the policy update of an agent from being too large, so that larger step sizes may be used without negatively affecting the gradient ascent process. === Basic concepts === To begin the PPO training process, the agent is set in an environment to perform actions based on its current input. In the early phase of training, the agent can freely explore solutions and keep track of the result. Later, with a certain amount of transition samples and policy updates, the agent will select an action to take by randomly sampling from the probability distribution P ( A | S ) {\displaystyle P(A|S)} generated by the policy network. The actions that are most likely to be beneficial will have the highest probability of being selected from the random sample. After an agent arrives at a different scenario (a new state) by acting, it is rewarded with a positive reward or a negative reward. The objective of an agent is to maximize the cumulative reward signal across sequences of states, known as episodes. === Policy gradient laws: the advantage function === The advantage function (denoted as A {\displaystyle A} ) is central to PPO, as it tries to answer the question of whether a specific action of the agent is better or worse than some other possible action in a given state. By definition, the advantage function is an estimate of the relative value for a selected action. If the output of this function is positive, it means that the action in question is better than the average return, so the possibilities of selecting that specific action will increase. The opposite is true for a negative advantage output. The advantage function can be defined as A = Q − V {\displaystyle A=Q-V} , where Q {\displaystyle Q} is the discounted sum of rewards (the total weighted reward for the completion of an episode) and V {\displaystyle V} is the baseline estimate. Since the advantage function is calculated after the completion of an episode, the program records the outcome of the episode. Therefore, calculating advantage is essentially an unsupervised learning problem. The baseline estimate comes from the value function that outputs the expected discounted sum of an episode starting from the current state. In the PPO algorithm, the baseline estimate will be noisy (with some variance), as it also uses a neural network, like the policy function itself. With Q {\displaystyle Q} and V {\displaystyle V} computed, the advantage function is calculated by subtracting the baseline estimate from the actual discounted return. If A > 0 {\displaystyle A>0} , the actual return of the action is better than the expected return from experience; if A < 0 {\displaystyle A<0} , the actual return is worse. === Ratio function === In PPO, the ratio function ( r t {\displaystyle r_{t}} ) calculates the probability of selecting action a {\displaystyle a} in state s {\displaystyle s} given the current policy network, divided by the previous probability under the old policy. In other words: If r t ( θ ) > 1 {\displaystyle r_{t}(\theta )>1} , where θ {\displaystyle \theta } are the policy network parameters, then selecting action a {\displaystyle a} in state s {\displaystyle s} is more likely based on the current policy than the previous policy. If 0 ≤ r t ( θ ) < 1 {\displaystyle 0\leq r_{t}(\theta )<1} , then selecting actio

Gamma (app)

Gamma is a web-based software platform that uses artificial intelligence to generate presentations, documents, webpages, and other visual content. The platform allow users to create structured layouts and draft text based on prompts or uploaded material. It operates as an online application and provides tools for editing, organizing, and sharing content. == History == Gamma was established in the early 2020s by Grant Lee, James Fox, and Jon Noronha during a period of increased development in artificial intelligence–based productivity software. The platform was introduced as a web-based format designed to present information through structured visual layouts rather than traditional slide-based presentations. Its interface was developed to adapt content to different screen sizes and devices. In later updates, Gamma expanded its functionality to support additional formats, including documents and simple webpages. By November 2025, the company reported that the platform had reached approximately 70 million users. Gamma has raised venture capital funding from a number of technology-focused investors since its founding. == Features == Gamma allows users to create presentations, documents, and webpages by entering prompts, pasting text, or uploading source files. The platform uses artificial intelligence to generate draft text, organize information, and apply structured layouts. Users can edit generated material manually and adjust formatting, structure, and visual elements. The software also supports collaborative editing, allowing multiple users to contribute to and revise the same project. Instead of relying only on fixed slide-based formats, Gamma presents content in scrollable layouts designed for web viewing across different screen sizes. Projects created on the platform can be shared through web links or exported to formats compatible with other software. Gamma also provides integration options and developer access through an application programming interface (API). == Technology == Gamma uses generative artificial intelligence models to interpret user input and generate structured content. The software automates elements of layout selection, formatting, and visual presentation. As with other AI-assisted tools, output produced by the system may require human review and revision to ensure accuracy and appropriate context. == Funding == Gamma has raised venture capital funding from a number of technology-focused investors since its founding. In November 2025, the company announced a Series B funding round that raised $68 million at a reported valuation of approximately $2.1 billion. Investors in the round included Andreessen Horowitz, Accel, and Uncork Capital, among others. == Controversy == In 2025, cybersecurity researchers reported that Gamma had been used in a phishing campaign targeting Microsoft accounts. Attackers shared links to presentations hosted on the platform that redirected users to a spoofed Microsoft SharePoint login page intended to collect credentials. Researchers noted that the incident reflected the broader misuse of legitimate online services in phishing schemes.

Cortica

Headquartered in Tel Aviv Cortica utilizes unsupervised learning methods to recognize and analyze digital images and video. The technology developed by the Cortica team is based on research of the function of the human brain. == Company Founding == Cortica was founded in 2007 by Igal Raichelgauz, Karina Odinaev and Yehoshua Zeevi. Together, the founders developed the company’s core technology while at Technion – Israel Institute of Technology. By combining discoveries in neuroscience with developments in computer programming, the team created technology that possesses the ability to interpret large amounts of visual data with increased accuracy. This technology, called Image2Text, is based on the founders’ work in digitally replicating cortical neural networks’ ability to identify complex patterns within massive quantities of ambiguous and noisy data. Cortica’s offerings have application in the automotive industry, media industries, as well as the smart city and medical industries. Industry experts suggest that the self-driving automotive industry alone will be worth upwards of $7 trillion while each connected car is expected to generate 4,000 GB of data per day. Beyond that, industry analysts expect the proliferation of surveillance cameras to continue leading to an expected 2,500 Petabytes of data being generated daily by new surveillance cameras. Cortica operates in these high scale industries. The company currently employs professionals from many domains including AI researchers as well as veterans of intelligence units within the Israeli Defense Forces. == Research and Technology == In 2006, Founders Raichelgauz, Odinaev, and Zeevi shared their findings with the 28th IEEE EMBS Annual International Conference in New York in a paper titled, “Natural Signal Classification by Neural Cliques and Phase-Locked Attractors”. That same year, the team also published “Cliques in Neural Ensembles as Perception Carriers" CB Insights recently identified Cortica as the number one patent holder among AI companies. Cortica is researching to develop a machine-learning driving system which can identify objects and pedestrians. Connecting to it, Elon Musk has been rumored to partner with Cortica for his electric car company, Tesla. However, Tesla denies it stating that Musk did not discuss a collaboration with artificial intelligence firm Cortica. == Funding == Cortica raised $7 million in its Series A funding round, announced in August 2012. Investors included Horizons Ventures (the investment firm of Hong Kong billionaire Li Ka-Shing), and Ynon Kreiz, the former chairman and CEO of the Endemol Group. In May 2013, it was announced that Cortica had raised $1.5 million from Russian firm Mail.ru Group. It later transpired that this was a part of Cortica's Series B funding round for $6.4 million, announced in June 2013. The round was led by Horizons Ventures, with participation from the Russian firm Mail.ru Group and other angel investors. In its fourth funding round, Cortica has raised $20 million, bringing the total investments to $38 million. According to a report from The Israeli lead Daily economic newspaper, TheMarker, the fourth round was led by a strategic Chinese investor who will probably help the company expand into the Asian market. == Media coverage == GigaOm listed Cortica as one of the top deep learning startups in a November 2013 article surveying the field, along with AlchemyAPI, Ersatz, and Semantria. Business Insider ranked Cortica as one of the coolest tech companies in Israel. CB Insights has identified Cortica as the top patent holding AI company. In 2017 several leading automotive media outlets covered the launch of Cortica's automotive business unit

Imageability

Imageability is a measure of how easily a physical object, word or environment will evoke a clear mental image in the mind of any person observing it. It is used in architecture and city planning, in psycholinguistics, and in automated computer vision research. In automated image recognition, training models to connect images with concepts that have low imageability can lead to biased and harmful results. == History and components == Kevin A. Lynch first introduced the term, "imageability" in his 1960 book, The Image of the City. In the book, Lynch argues cities contain a key set of physical elements that people use to understand the environment, orient themselves inside of it, and assign it meaning. Lynch argues the five key elements that impact the imageability of a city are Paths, Edges, Districts, Nodes, and Landmarks. Paths: channels in which people travel. Examples: streets, sidewalks, trails, canals, railroads. Edges: objects that form boundaries around space. Examples: walls, buildings, shoreline, curbstone, streets, and overpasses. Districts: medium to large areas people can enter into and out of that have a common set of identifiable characteristics. Nodes: large areas people can enter, that serve as the foci of the city, neighborhood, district, etc. Landmarks: memorable points of reference people cannot enter into. Examples: signs, mountains and public art. In 1914, half a century before The Image of the City was published, Paul Stern discussed a concept similar to imageability in the context of art. Stern, in Susan Langer's Reflections on Art, names the attribute that describes how vividly and intensely an artistic object could be experienced apparency. == In computer vision == Automated image recognition was developed by using machine learning to find patterns in large, annotated datasets of photographs, like ImageNet. Images in ImageNet are labelled using concepts in WordNet. Concepts that are easily expressed verbally, like "early", are seen as less "imageable" than nouns referring to physical objects like "leaf". Training AI models to associate concepts with low imageability with specific images can lead to problematic bias in image recognition algorithms. This has particularly been critiqued as it relates to the "person" category of WordNet and therefore also ImageNet. Trevor Pagan and Kate Crawford demonstrated in their essay "Excavating AI" and their art project ImageNet Roulette how this leads to photos of ordinary people being labelled by AI systems as "terrorists" or "sex offenders". Images in datasets are often labelled as having a certain level of imageability. As described by Kaiyu Yang, Fei-Fei Li and co-authors, this is often done following criteria from Allan Paivio and collaborators' 1968 psycholinguistic study of nouns. Yang el.al. write that dataset annotators tasked with labelling imageability "see a list of words and rate each word on a 1-7 scale from 'low imagery' to 'high imagery'. To avoid biased or harmful image recognition and image generation, Yang et.al. recommend not training vision recognition models on concepts with low imageability, especially when the concepts are offensive (such as sexual or racial slurs) or sensitive (their examples for this category include "orphan", "separatist", "Anglo-Saxon" and "crossover voter"). Even "safe" concepts with low imageability, like "great-niece" or "vegetarian" can lead to misleading results and should be avoided.

Chainer

Chainer is an open source deep learning framework written purely in Python on top of NumPy and CuPy Python libraries. The development is led by Japanese venture company Preferred Networks in partnership with IBM, Intel, Microsoft, and Nvidia. Chainer is notable for its early adoption of "define-by-run" scheme, as well as its performance on large scale systems. The first version was released in June 2015 and has gained large popularity in Japan since then. Furthermore, in 2017, it was listed by KDnuggets in top 10 open source machine learning Python projects. In December 2019, Preferred Networks announced the transition of its development effort from Chainer to PyTorch and it will only provide maintenance patches after releasing v7. == Define-by-run == Chainer was the first deep learning framework to introduce the define-by-run approach. The traditional procedure to train a network was in two phases: define the fixed connections between mathematical operations (such as matrix multiplication and nonlinear activations) in the network, and then run the actual training calculation. This is called the define-and-run or static-graph approach. Theano and TensorFlow are among the notable frameworks that took this approach. In contrast, in the define-by-run or dynamic-graph approach, the connection in a network is not determined when the training is started. The network is determined during the training as the actual calculation is performed. One of the advantages of this approach is that it is intuitive and flexible. If the network has complicated control flows such as conditionals and loops, in the define-and-run approach, specially designed operations for such constructs are needed. On the other hand, in the define-by-run approach, programming language's native constructs such as if statements and for loops can be used to describe such flow. This flexibility is especially useful to implement recurrent neural networks. Another advantage is ease of debugging. In the define-and-run approach, if an error (such as numeric error) has occurred in the training calculation, it is often difficult to inspect the fault, because the code written to define the network and the actual place of the error are separated. In the define-by-run approach, you can just suspend the calculation with the language's built-in debugger and inspect the data that flows on your code of the network. Define-by-run has gained popularity since the introduction by Chainer and is now implemented in many other frameworks, including PyTorch and TensorFlow. == Extension libraries == Chainer has four extension libraries, ChainerMN, ChainerRL, ChainerCV and ChainerUI. ChainerMN enables Chainer to be used on multiple GPUs with performance significantly faster than other deep learning frameworks. A supercomputer running Chainer on 1024 GPUs processed 90 epochs of ImageNet dataset on ResNet-50 network in 15 minutes, which is four times faster than the previous record held by Facebook. ChainerRL adds state of art deep reinforcement learning algorithms, and ChainerUI is a management and visualization tool. == Applications == Chainer is used as the framework for PaintsChainer, a service which does automatic colorization of black and white, line only, draft drawings with minimal user input.

Micro stuttering

Micro stuttering is a visual artifact in real-time computer graphics in which the time intervals between consecutively displayed frames are uneven, even though the average frame rate reported by benchmarking software appears adequate. Tools such as 3DMark typically compute frame rates over intervals of one second or more, which can conceal momentary drops in the instantaneous frame rate that the viewer perceives as hitching or jerking of on-screen motion. At low frame rates the effect is visible as a stutter in moving images, degrading the experience in interactive applications such as video games. In severe cases a lower but more consistent frame rate can appear smoother than a higher but more erratic one. The term gained prominence in the late 2000s in discussions of multi-GPU rendering (see History), but micro stuttering also affects single-GPU systems. Common causes on modern hardware include real-time shader compilation, asset streaming from storage, VRAM exhaustion, and driver bugs. == Causes == === Shader compilation === A common cause of micro stuttering on modern PCs is real-time shader compilation. Shaders are small programs that instruct the GPU on how to render visual effects such as lighting, shadows, and reflections. On consoles, developers can pre-compile all shaders for the known, fixed hardware. On PCs, the variety of GPU architectures means shaders must often be compiled at run time, either when the game launches or during gameplay itself. When the rendering engine encounters a shader that has not yet been compiled, the CPU must finish the compilation before the GPU can draw the affected object. This causes a spike in frame time that the player perceives as a hitch. The problem has been particularly associated with games built on Unreal Engine 4 running under DirectX 12, because DX12 shifts more shader management responsibility to the application. Several techniques exist to reduce shader compilation stutter. Pipeline State Object (PSO) pre-caching records the shader permutations used at runtime so that they can be compiled in advance on subsequent launches. Asynchronous shader compilation moves the work to background CPU threads to avoid blocking the main rendering thread. Platform-level services such as Steam's shader pre-caching distribute previously compiled shaders to users with matching GPU hardware. The Steam Deck, which contains a single fixed GPU, benefits from pre-compiled shader caches because all units share the same hardware configuration. === Other causes === Micro stuttering on single-GPU systems can have several additional causes. CPU bottlenecks or scheduling interruptions from background tasks can prevent the processor from preparing frames at regular intervals. Asset streaming during gameplay (loading textures, geometry, or audio from storage) can produce hitches sometimes called traversal stutter; the use of solid-state drives and technologies such as DirectStorage has reduced but not eliminated this. VRAM exhaustion forces data to be swapped between video memory and system memory over the PCI Express bus, which is slower. Graphics driver bugs can also introduce stutter; Nvidia released hotfix driver 551.46 in February 2024 to correct intermittent micro stuttering when V-Sync was enabled. == Measurement == Micro stuttering drew attention to the limitations of average frame rate as a performance metric. In 2013, Scott Wasson at The Tech Report published a series of articles advocating frame time analysis, in which the delivery time of every individual frame is recorded and plotted rather than collapsed into a single frames-per-second figure. This approach was adopted by other hardware review publications in the following years. GPU reviews now routinely report 1% low and 0.1% low frame rates alongside the average. The 1% low is the average frame rate of the slowest 1% of frames in a sample; it serves as an indicator of worst-case smoothness. A large gap between the average and the 1% low suggests poor frame pacing. Tools for capturing per-frame timing data include FRAPS, PresentMon, OCAT, CapFrameX, and MSI Afterburner with RivaTuner Statistics Server. == Mitigation == === Frame pacing === Frame pacing is a software technique that regulates the timing of frame delivery to produce even intervals between displayed frames. Game engines, GPU drivers, and platform libraries all implement frame pacing strategies to varying degrees. On mobile platforms, Google provides the Android Frame Pacing library (Swappy) as part of the Android Game Development Kit. In December 2025, the Khronos Group published the VK_EXT_present_timing Vulkan extension, giving developers explicit control over presentation timing in a cross-platform graphics API for the first time. === Variable refresh rate === Variable refresh rate (VRR) display technologies allow a monitor's refresh rate to change to match the GPU's frame output. Implementations include Nvidia G-Sync (2013), AMD FreeSync (2015), and the VESA Adaptive-Sync standard built into DisplayPort 1.2a and later. VRR eliminates the screen tearing that results from a mismatch between frame rate and refresh rate, and avoids the frame-holding behaviour of V-Sync that can itself cause stutter. It is effective at smoothing moderate frame rate fluctuations but cannot compensate for large sudden spikes in frame time such as those caused by shader compilation or heavy asset streaming. VRR support has become standard in gaming monitors, televisions (via HDMI 2.1), and the Xbox Series X/S and PlayStation 5 consoles. === Frame generation === Beginning with DLSS 3 on the GeForce RTX 40 series in 2022, Nvidia introduced AI-based frame generation, which uses dedicated optical flow hardware and a neural network to create new frames between traditionally rendered ones. AMD followed with FSR 3 in 2023, using an algorithmic approach, and the AI-based FSR 4 for the Radeon RX 9000 series in 2025. DLSS 4, released in January 2025 for the GeForce RTX 50 series, can generate up to three frames per rendered frame using a technique called Multi Frame Generation. Frame generation increases the displayed frame rate but introduces its own frame pacing concerns. If the underlying rendered frames are unevenly timed, the interpolated frames can make the unevenness more apparent rather than less. DLSS 4 addresses this with hardware-level flip metering on the GPU's display engine, which controls the timing of frame presentation more precisely than the CPU-based pacing used in DLSS 3. Both vendors pair frame generation with latency-reduction features (Nvidia Reflex and AMD Anti-Lag+) to offset the additional input latency that results from inserting synthetic frames into the pipeline. === Frame rate limiters === Capping the frame rate below the display's maximum refresh rate, using tools such as RivaTuner Statistics Server, in-game limiters, or driver-level settings, is a common way to improve frame pacing. Preventing the GPU from running ahead of the display reduces variability in frame delivery times and can produce a smoother result than an uncapped but more irregular frame rate. == History == === Multi-GPU configurations === Micro stuttering was first widely documented in the late 2000s as a side effect of multi-GPU configurations using Alternate Frame Rendering (AFR), in which consecutive frames are assigned to alternating GPUs. Because each GPU may take a different amount of time to complete its assigned frame — due to varying scene complexity, driver scheduling, or inter-GPU communication overhead — the resulting frame delivery is irregular even when the average frame rate is high. Both Nvidia SLI and AMD CrossFireX were affected, with dual-GPU setups exhibiting the worst frame pacing irregularities. In 2012 benchmarks using Battlefield 3, dual Radeon HD 7970 cards in CrossFire showed 85% variation in frame delivery times compared with 7% for a single card, while dual GeForce GTX 680 cards in SLI showed only 7% variation compared with 5% for a single card. Multi-GPU micro stuttering became a significant factor in the eventual decline and discontinuation of consumer multi-GPU gaming. Nvidia restricted SLI to a handful of enthusiast-class cards from the GeForce 10 series onward, then replaced it with NVLink on the GeForce RTX 20 series, which saw limited gaming adoption. AMD ceased active CrossFire development around 2017. By the mid-2020s, neither vendor's current consumer GPUs support multi-GPU rendering for games. Other factors that contributed to the decline include DirectX 12 placing multi-GPU support in the hands of game developers rather than driver authors, the incompatibility of temporal anti-aliasing and other temporal rendering techniques with AFR, and the increasing size, power draw, and cost of individual GPUs. The third-party utility RadeonPro could reduce CrossFire micro stuttering through dynamic V-Sync and frame pacing adjustments, and AMD later introduced a driver-level frame paci

Open Neural Network Exchange

The Open Neural Network Exchange (ONNX) [ˈɒnɪks] is an open-source artificial intelligence ecosystem of technology companies and research organizations that establish open standards for representing machine learning algorithms and software tools to enable a standard format for representing machine learning models. ONNX is available on GitHub. == History == ONNX was originally named Toffee and was developed by the PyTorch team at Facebook. In September 2017 it was renamed to ONNX and announced by Facebook and Microsoft. Later, IBM, Huawei, Intel, AMD, Arm and Qualcomm announced support for the initiative. In October 2017, Microsoft announced that it would add its Cognitive Toolkit and Project Brainwave platform to the initiative. In November 2019 ONNX was accepted as graduate project in Linux Foundation AI. In October 2020 Zetane Systems became a member of the ONNX ecosystem. == Intent == The initiative targets: === Framework interoperability === Enable developers to move machine learning models between different frameworks, which may be used at different stages of the development process, such as training, architecture design, or deployment on mobile devices. === Shared optimization === Provide a common representation that can be used by hardware vendors and other developers to apply optimizations to artificial neural network models across multiple machine learning frameworks. == Contents == ONNX provides definitions of an extensible computation graph model, built-in operators and standard data types, focused on inferencing (evaluation).. The container format is Protocol Buffers. Each computation dataflow graph is a list of nodes that form an acyclic graph. Nodes have inputs and outputs. Each node is a call to an operator. Metadata documents the graph. Built-in operators are to be available on each ONNX-supporting framework. ONNX models can be trained in a single framework, such as PyTorch or TensorFlow, and then exported to ONNX. This format allows models to be transferred from the training framework to other environments for testing or deployment. Once a model is in ONNX format, it can be executed in different runtime systems or on various hardware platforms, such as GPUs or specialized AI accelerators. Using a common format enables the same model representation to be used across multiple systems and frameworks.