Zotero Catalogue

E-waste contaminates soil and water, threatening agriculture and human health. Learn how proper recycling and stricter regulations can mitigate its impact.

Artificial intelligence has been successfully applied in various fields, one of which is computer vision. In this study, a deep neural network (DNN) was adopted for Facial emotion recognition (FER). One of the objectives in this study is to identify the critical facial features on which the DNN model focuses for FER. In particular, we utilized a convolutional neural network (CNN), the combination of squeeze-and-excitation network and the residual neural network, for the task of FER. We utilized AffectNet and the Real-World Affective Faces Database (RAF-DB) as the facial expression databases that provide learning samples for the CNN. The feature maps were extracted from the residual blocks for further analysis. Our analysis shows that the features around the nose and mouth are critical facial landmarks for the neural networks. Cross-database validations were conducted between the databases. The network model trained on AffectNet achieved 77.37% accuracy when validated on the RAF-DB, while the network model pretrained on AffectNet and then transfer learned on the RAF-DB results in validation accuracy of 83.37%. The outcomes of this study would improve the understanding of neural networks and assist with improving computer vision accuracy.

Despite the recent progresses, particularly in developing Language Models, there are fundamental challenges and unanswered questions about how such models can continually learn/memorize, self-improve, and find effective solutions. In this paper, we present a new learning paradigm, called Nested Learning (NL), that coherently represents a machine learning model with a set of nested, multi-level, and/or parallel optimization problems, each of which with its own context flow. Through the lenses of NL, existing deep learning methods learns from data through compressing their own context flow, and in-context learning naturally emerges in large models. NL suggests a philosophy to design more expressive learning algorithms with more levels, resulting in higher-order in-context learning and potentially unlocking effective continual learning capabilities. We advocate for NL by presenting three core contributions: (1) Expressive Optimizers: We show that known gradient-based optimizers, such as Adam, SGD with Momentum, etc., are in fact associative memory modules that aim to compress the gradients' information (by gradient descent). Building on this insight, we present other more expressive optimizers with deep memory and/or more powerful learning rules; (2) Self-Modifying Learning Module: Taking advantage of NL's insights on learning algorithms, we present a sequence model that learns how to modify itself by learning its own update algorithm; and (3) Continuum Memory System: We present a new formulation for memory system that generalizes the traditional viewpoint of long/short-term memory. Combining our self-modifying sequence model with the continuum memory system, we present a continual learning module, called Hope, showing promising results in language modeling, knowledge incorporation, and few-shot generalization tasks, continual learning, and long-context reasoning tasks.

Scaling models has led to significant advancements in deep learning, but training these models in decentralized settings remains challenging due to communication bottlenecks. While existing compression techniques are effective in data-parallel, they do not extend to model parallelism. Unlike data-parallel training, where weight gradients are exchanged, model-parallel requires compressing activations and activation gradients as they propagate through layers, accumulating compression errors. We propose a novel compression algorithm that compresses both forward and backward passes, enabling up to 99% compression with no convergence degradation with negligible memory/compute overhead. By leveraging a recursive structure in transformer networks, we predefine a low-dimensional subspace to confine the activations and gradients, allowing full reconstruction in subsequent layers. Our method achieves up to 100x improvement in communication efficiency and enables training billion-parameter-scale models over low-end GPUs connected via consumer-grade internet speeds as low as 80Mbps, matching the convergence of centralized datacenter systems with 100Gbps connections with model parallel.

Generative models trained on internet data have revolutionized how text, image, and video content can be created. Perhaps the next milestone for generative models is to simulate realistic experience in response to actions taken by humans, robots, and other interactive agents. Applications of a real-world simulator range from controllable content creation in games and movies, to training embodied agents purely in simulation that can be directly deployed in the real world. We explore the possibility of learning a universal simulator (UniSim) of real-world interaction through generative modeling. We first make the important observation that natural datasets available for learning a real-world simulator are often rich along different dimensions (e.g., abundant objects in image data, densely sampled actions in robotics data, and diverse movements in navigation data). With careful orchestration of diverse datasets, each providing a different aspect of the overall experience, we can simulate the visual outcome of both high-level instructions such as "open the drawer" and low-level controls from otherwise static scenes and objects. We use the simulator to train both high-level vision-language policies and low-level reinforcement learning policies, each of which can be deployed in the real world in zero shot after training purely in simulation. We also show that other types of intelligence such as video captioning models can benefit from training with simulated experience, opening up even wider applications. Video demos can be found at https://universal-simulator.github.io.

How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? This position paper proposes an architecture and training paradigms with which to construct autonomous intelligent agents. It combines concepts such as configurable predictive world model, behavior driven through intrinsic motivation, and hierarchical joint embedding architectures trained with self-supervised learning.

Autonomous driving promises transformative improvements to transportation, but building systems capable of safely navigating the unstructured complexity of real-world scenarios remains challenging. A critical problem lies in effectively predicting the various potential outcomes that may emerge in response to the vehicle's actions as the world evolves. To address this challenge, we introduce GAIA-1 ('Generative AI for Autonomy'), a generative world model that leverages video, text, and action inputs to generate realistic driving scenarios while offering fine-grained control over ego-vehicle behavior and scene features. Our approach casts world modeling as an unsupervised sequence modeling problem by mapping the inputs to discrete tokens, and predicting the next token in the sequence. Emerging properties from our model include learning high-level structures and scene dynamics, contextual awareness, generalization, and understanding of geometry. The power of GAIA-1's learned representation that captures expectations of future events, combined with its ability to generate realistic samples, provides new possibilities for innovation in the field of autonomy, enabling enhanced and accelerated training of autonomous driving technology.

We explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted from the world model as inputs to an agent, we can train a very compact and simple policy that can solve the required task. We can even train our agent entirely inside of its own hallucinated dream generated by its world model, and transfer this policy back into the actual environment. An interactive version of this paper is available at https://worldmodels.github.io/

Developing a general algorithm that learns to solve tasks across a wide range of applications has been a fundamental challenge in artificial intelligence. Although current reinforcement learning algorithms can be readily applied to tasks similar to what they have been developed for, configuring them for new application domains requires significant human expertise and experimentation. We present DreamerV3, a general algorithm that outperforms specialized methods across over 150 diverse tasks, with a single configuration. Dreamer learns a model of the environment and improves its behavior by imagining future scenarios. Robustness techniques based on normalization, balancing, and transformations enable stable learning across domains. Applied out of the box, Dreamer is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. This achievement has been posed as a significant challenge in artificial intelligence that requires exploring farsighted strategies from pixels and sparse rewards in an open world. Our work allows solving challenging control problems without extensive experimentation, making reinforcement learning broadly applicable.

Language models show a surprising range of capabilities, but the source of their apparent competence is unclear. Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state. Interventional experiments indicate this representation can be used to control the output of the network and create "latent saliency maps" that can help explain predictions in human terms.

Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents--computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent's experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors: for example, starting with only a single user-specified notion that one agent wants to throw a Valentine's Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture--observation, planning, and reflection--each contribute critically to the believability of agent behavior. By fusing large language models with computational, interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior.

We study allowing large language models (LLMs) to process arbitrarily long prompts through the lens of inference-time scaling. We propose Recursive Language Models (RLMs), a general inference strategy that treats long prompts as part of an external environment and allows the LLM to programmatically examine, decompose, and recursively call itself over snippets of the prompt. We find that RLMs successfully handle inputs up to two orders of magnitude beyond model context windows and, even for shorter prompts, dramatically outperform the quality of base LLMs and common long-context scaffolds across four diverse long-context tasks, while having comparable (or cheaper) cost per query.

Document preview for Milieudefensie et al. v. Royal Dutch Shell plc. - appeal

Genetically, I should be bald.    I started to lose my hair and go gray in my late 20s. Now, at 46, I’ve got a full head of hair and ~50% of my gray is gone. Here’s how I did it:   Start early and be proactive - I made the mistake of addressing my hair loss and graying after noticing it. By age 20, about 20% of men a

We introduce Genie, the first generative interactive environment trained in an unsupervised manner from unlabelled Internet videos. The model can be prompted to generate an endless variety of action-controllable virtual worlds described through text, synthetic images, photographs, and even sketches. At 11B parameters, Genie can be considered a foundation world model. It is comprised of a spatiotemporal video tokenizer, an autoregressive dynamics model, and a simple and scalable latent action model. Genie enables users to act in the generated environments on a frame-by-frame basis despite training without any ground-truth action labels or other domain-specific requirements typically found in the world model literature. Further the resulting learned latent action space facilitates training agents to imitate behaviors from unseen videos, opening the path for training generalist agents of the future.

Biological brains demonstrate complex neural activity, where neural dynamics are critical to how brains process information. Most artificial neural networks ignore the complexity of individual neurons. We challenge that paradigm. By incorporating neuron-level processing and synchronization, we reintroduce neural timing as a foundational element. We present the Continuous Thought Machine (CTM), a model designed to leverage neural dynamics as its core representation. The CTM has two innovations: (1) neuron-level temporal processing, where each neuron uses unique weight parameters to process incoming histories; and (2) neural synchronization as a latent representation. The CTM aims to strike a balance between neuron abstractions and biological realism. It operates at a level of abstraction that effectively captures essential temporal dynamics while remaining computationally tractable. We demonstrate the CTM's performance and versatility across a range of tasks, including solving 2D mazes, ImageNet-1K classification, parity computation, and more. Beyond displaying rich internal representations and offering a natural avenue for interpretation owing to its internal process, the CTM is able to perform tasks that require complex sequential reasoning. The CTM can also leverage adaptive compute, where it can stop earlier for simpler tasks, or keep computing when faced with more challenging instances. The goal of this work is to share the CTM and its associated innovations, rather than pushing for new state-of-the-art results. To that end, we believe the CTM represents a significant step toward developing more biologically plausible and powerful artificial intelligence systems. We provide an accompanying interactive online demonstration at https://pub.sakana.ai/ctm/ and an extended technical report at https://pub.sakana.ai/ctm/paper .

A vocal technique reference covering 21 techniques: registers, styles, effects, embellish, and dynamics.

Why I stopped reading code and started watching it stream by.

This is a course on the theory of quantum computing. It consists of 16 lessons, each with a video and written component, covering the basics of quantum information, quantum algorithms (including query algorithms, Shor's algorithm for integer factorization, and Grover's algorithm), the general formulation of quantum information (including density matrices, quantum channels, and general measurements), and quantum error correction (including the basics, the stabilizer formalism, CSS codes, the toric code, and fault-tolerant quantum computation).

Justine Musk is a Canadian author and ex-wife to Elon Musk.

Learn how to write killer sales emails at scale that engage buyers and drive meetings — including best practices for relevance, structure, concise messaging, Calls to Action (CTAs), and scaling outbound outreach without sacrificing quality.

What we learned from three iterations of a performance engineering take-home that Claude keeps beating.

Diffusion language models hold the promise of fast parallel generation, while autoregressive (AR) models typically excel in quality due to their causal structure aligning naturally with language modeling. This raises a fundamental question: can we achieve a synergy with high throughput, higher GPU utilization, and AR level quality? Existing methods fail to effectively balance these two aspects, either prioritizing AR using a weaker model for sequential drafting (speculative decoding), leading to lower drafting efficiency, or using some form of left-to-right (AR-like) decoding logic for diffusion, which still suffers from quality degradation and forfeits its potential parallelizability. We introduce TiDAR, a sequence-level hybrid architecture that drafts tokens (Thinking) in Diffusion and samples final outputs (Talking) AutoRegressively - all within a single forward pass using specially designed structured attention masks. This design exploits the free GPU compute density, achieving a strong balance between drafting and verification capacity. Moreover, TiDAR is designed to be serving-friendly (low overhead) as a standalone model. We extensively evaluate TiDAR against AR models, speculative decoding, and diffusion variants across generative and likelihood tasks at 1.5B and 8B scales. Thanks to the parallel drafting and sampling as well as exact KV cache support, TiDAR outperforms speculative decoding in measured throughput and surpasses diffusion models like Dream and Llada in both efficiency and quality. Most notably, TiDAR is the first architecture to close the quality gap with AR models while delivering 4.71x to 5.91x more tokens per second.

Get 25% off Blinkist Annual premium! Start your 7-day free trial by clicking here: http://blinkist.com/machinelearningst... This is the first part of the Yannic Kilcher interview we posted recently! LAION petition mentioned: https://www.openpetition.eu/petition/... Yannic's channel: ‪@YannicKilcher‬ Tim Scarfe and Yannic Kilcher engaged in a wide-ranging discussion on the rise of powerful language models like ChatGPT and GPT-4 and their implications. While they believe these models will significantly transform industries and jobs, they do not foresee mass unemployment or existential risks from superintelligence. Scarfe and Kilcher think most people will adapt to work with AI, as they did with smartphones. The key will be learning to effectively apply these new tools. Regarding concerns about bias and misinformation, they acknowledge these are issues but not fundamentally new problems - people have always spread misinformation, and we must develop better digital literacy and detection methods. Regulating the technology itself likely will not solve underlying social challenges. Though skeptical of superintelligence, Scarfe and Kilcher expect models will become very capable in narrow domains, sometimes in ways beyond human understanding. But human judgment will remain critical to properly apply and constrain them. They believe progress should be encouraged if we are thoughtful and deliberate. Overall benefits likely far outweigh risks. Rather than limiting access, Scarfe and Kilcher support distributing models widely so researchers and users can better understand, improve and constrain them. Regulation and oversight still matter, but restricting progress seems misguided and unlikely to succeed. While powerful models may transform our lives, Scarfe and Kilcher expect human creativity and values to endure. With prudent development and policymaking, language models could usher in a new era of possibility. But we must be vigilant and deliberate to ensure the responsible and equitable progress of this technology. Overall, Scarfe and Kilcher express optimistic realism about the rise of language models. With open and proactively guided progress, they foresee emerging capabilities enhancing rather than diminishing human potential. But we must be willing partners in developing and applying these tools, not hapless bystanders. In shaping our future with AI, our values and judgment will be vital - these technologies remain a means, not an end. Their promise depends on human wisdom and oversight. TOC 0:00:00 - Introduction 0:03:09 - Background and Motivation 0:03:29 - AI Frameworks and Approaches 0:06:05 - Impact of Language Models 0:18:54 - Misinformation and Autonomy 0:29:43 - Superintelligence and AI Risk 0:45:15 - Discussion on the fear of superintelligence 0:51:37 - Understanding GPT-4 and its capabilities 0:55:55 - Comparing AI systems to human intelligence

Manage your credits and payment history

Ryan LaRose, Andrea Mari, Sarah Kaiser, Peter J. Karalekas, Andre A. Alves, Piotr Czarnik, Mohamed El Mandouh, Max H. Gordon, Yousef Hindy, Aaron Robertson, Purva Thakre, Misty Wahl, Danny Samuel, Rahul Mistri, Maxime Tremblay, Nick Gardner, Nathaniel T. Stemen, Nathan Shammah, and William J. Zeng, Quantum 6, 774 (2022). We introduce Mitiq, a Python package for error mitigation on noisy quantum computers. Error mitigation techniques can reduce the impact of noise on near-term quantum computers with minimal o…

The state of a physical system is the mathematical object that provides a complete information on the system. The knowledge of the state is equivalent to know the result of any possible measurement on the system. This chapter reviews quantum state estimation for a generic quantum system by quantum tomography i.e. from the measurement of a suitable set of observables, a quorum, on repeated preparations of the system. Topics include characterization of quora, determination of the expectation value of any operator (including the nondiagonal projectors needed to construct a matrix representation of the density operator), evaluation of pattern functions, effect of instrumental noise, and example of tomographic procedure for harmonic systems and spins.

The state of a physical system is the mathematical description of our knowledge of it, and provides information on its future and past. A state estimation technique is a method that provides the complete description of a system, i.e achieves the maximum possible knowledge of the state, thus allowing one to make the best, at least the best probabilistic, predictions on the results of any measurement that may be performed on the system.

Maximum Likelihood estimation is a versatile tool covering wide range of applications, but its benefits are apparent particularly in the quantum domain. For a given set of measurements, the most likely state is estimated. Though this problem is nonlinear, it can be effectively solved by an iterative algorithm exploiting the convexity of the likelihood functional and the manifold of density matrices. This formulation fully replaces the inverse Radon transformation routinely used for tomographic reconstructions. Moreover, it provides the most efficient estimation strategy saturating the Cramer-Rao lower bound asymptotically. In this sense it exploits the acquired data set in the optimal way and minimizes the artifacts associated with the reconstruction procedure. The idea of maximum likelihood reconstruction is further extended to the estimation of quantum processes, measurements, and discrimination between quantum states. This technique is well suited for future applications in quantum information science due to its ability to quantify very subtle and fragile quantum effects.

Maximum Likelihood estimation is a versatile tool covering wide range of applications, but its benefits are apparent particularly in the quantum domain. For a given set of measurements, the most likely state is estimated. Though this problem is nonlinear, it can be effectively solved by an iterative algorithm exploiting the convexity of the likelihood functional and the manifold of density matrices. This formulation fully replaces the inverse Radon transformation routinely used for tomographic reconstructions. Moreover, it provides the most efficient estimation strategy saturating the Cramer-Rao lower bound asymptotically. In this sense it exploits the acquired data set in the optimal way and minimizes the artifacts associated with the reconstruction procedure. The idea of maximum likelihood reconstruction is further extended to the estimation of quantum processes, measurements, and discrimination between quantum states. This technique is well suited for future applications in quantum information science due to its ability to quantify very subtle and fragile quantum effects.

Dr. Crystal Senko Professor, Department of Physics and Astronomy Faculty of Science > Institute for Quantum Computing > Co-founder, Open Quantum Design Researchers from the University of Waterloo’s Faculty of Science and the Institute for Quantum Computing (IQC) are prioritizing collaboration over competition to advance quantum computer development and the field of quantum

Quantum computing has gained significant attention in recent years, with numerous algorithms and applications under active development. Limited by the current quantum technology, quantum noise and readout error have become critical issues. Various methods have been proposed to address readout error through error mitigation techniques, typically involving post-processing of measurement data. However, most of these methods increase the quantum hardware overhead, leading to higher computational costs. In this work, we present a machine-learning-based approach that minimizes hardware overhead while improving accuracy of the measurement probability distributions. We employed a convolutional neural network (CNN) autoencoder, commonly used for image denoising, as our baseline model. The datasets were derived from 4-qubit random circuits with depths ranging from 1 to 18, generated using Qiskit backends for target and noisy measurement data. The model was trained using mean squared error (MSE) as the loss function and Adam optimizer over 500 epochs, achieving an average noise reduction by 95% across the validation set, with no signs of overfitting. To validate the model's effectiveness across diverse quantum states, we conducted extensive tests on both typical quantum circuits and algorithms, including Grover's search algorithm, Quantum Fourier Transform, Haar random circuits and Trivial Paramagnet. The results demonstrated consistent and robust denoising in noisy measurement data, indicating that the autoencoder model is well-suited for efficient quantum error mitigation for current noisy quantum computers. This work contributes to the advancement of quantum error mitigation techniques using machine learning.

The matter distribution of the Universe can be mapped through the weak gravitational lensing (WL) effect: small distortions of the shapes of distant galaxies, which reflects the inhomogeneity of the cosmic density field. The most dominant contaminant in the WL effect is the shape noise; the signal is diluted due to the finite number of source galaxies. In order to explore the full potential of WL measurements, sharpening the signal by removing the shape noise from the observational data, i.e., WL denoising, is a pressing issue. Machine learning approaches, in particular, deep generative models, have proven effective at the WL denoising task. We implement a denoising model based on the diffusion model (DM) and conduct systematic in-depth comparisons with generative adversarial networks (GANs), which have been applied in previous works for WL denoising. Utilizing the large suite of mock simulations of WL observations, we demonstrate that DM surpasses GAN in the WL denosing task in multiple aspects: (1) the training process is more stable, (2) taking the average of multiple samples from DM can robustly reproduce the true signal, and (3) DM can recover various statistics with higher accuracy.

Traditional design of Laser Pulse Shapes (LPs) for Inertial Confinement Fusion (ICF) is a significant bottleneck, relying on computationally expensive simulations and manual iterative refinement. We introduce Diffusion-Driven Inverse Modeling for Laser Pulse Design (DiffICF), a generative inverse model that directly maps specified implosion outcomes to tailored LPs. DiffICF incorporates a physics-informed loss function that enforces known experimental and physical constraints. Moreover, it enables fine-grained control over pulse characteristics through constraint conditioning and inpainting. The efficacy of this framework was experimentally validated for optimizing implosion outcomes, offering a scalable, data-driven design tool to accelerate progress in fusion energy.

Learning an energy-based model (EBM) in the latent space of a top-down generative model offers a powerful framework for generation across many data modalities. However, it remains unclear how its interpretability can be used to guide model design, improve generative quality, and reduce training time. Moreover, the reliance on Langevin Monte Carlo (LMC) sampling presents challenges in efficiency and sampling multimodal latent distributions. We propose a novel adaptation of the Kolmogorov-Arnold representation theorem for generative modeling and introduce the Kolmogorov-Arnold Energy Model (KAEM) to take advantage of structural and inductive biases. By constraining the prior to univariate relationships, KAEM enables fast and exact inference via the inverse transform method. With the low dimensionality of the latent space and suitable inductive biases encoded, we demonstrate that importance sampling (IS) becomes a viable, unbiased, and highly efficient posterior sampler. For domains where IS fails, we introduce a strategy based on population-based LMC, decomposing the posterior into a sequence of annealed distributions to improve LMC mixing. KAEM balances common generative modeling trade-offs, offering fast inference, interpretability, and stable training, while being naturally suited to Zettascale Computing hardware.

The accumulation of physical errors1–3 prevents the execution of large-scale algorithms in current quantum computers. Quantum error correction4 promises a solution by encoding k logical qubits onto a larger number n of physical qubits, such that the physical errors are suppressed enough to allow running a desired computation with tolerable fidelity. Quantum error correction becomes practically realizable once the physical error rate is below a threshold value that depends on the choice of quantum code, syndrome measurement circuit and decoding algorithm5. We present an end-to-end quantum error correction protocol that implements fault-tolerant memory on the basis of a family of low-density parity-check codes6. Our approach achieves an error threshold of 0.7% for the standard circuit-based noise model, on par with the surface code7–10 that for 20 years was the leading code in terms of error threshold. The syndrome measurement cycle for a length-n code in our family requires n ancillary qubits and a depth-8 circuit with CNOT gates, qubit initializations and measurements. The required qubit connectivity is a degree-6 graph composed of two edge-disjoint planar subgraphs. In particular, we show that 12 logical qubits can be preserved for nearly 1 million syndrome cycles using 288 physical qubits in total, assuming the physical error rate of 0.1%, whereas the surface code would require nearly 3,000 physical qubits to achieve said performance. Our findings bring demonstrations of a low-overhead fault-tolerant quantum memory within the reach of near-term quantum processors.