Publications

Wuerstchen: An Efficient Architecture for Large-Scale Text-to-Image Diffusion Models
Pablo Pernias
Dominic Rampas
Mats Leon Richter
Christopher Pal
Marc Aubreville
We introduce W\"urstchen, a novel architecture for text-to-image synthesis that combines competitive performance with unprecedented cost-eff… (see more)ectiveness for large-scale text-to-image diffusion models. A key contribution of our work is to develop a latent diffusion technique in which we learn a detailed but extremely compact semantic image representation used to guide the diffusion process. This highly compressed representation of an image provides much more detailed guidance compared to latent representations of language and this significantly reduces the computational requirements to achieve state-of-the-art results. Our approach also improves the quality of text-conditioned image generation based on our user preference study. The training requirements of our approach consists of 24,602 A100-GPU hours - compared to Stable Diffusion 2.1's 200,000 GPU hours. Our approach also requires less training data to achieve these results. Furthermore, our compact latent representations allows us to perform inference over twice as fast, slashing the usual costs and carbon footprint of a state-of-the-art (SOTA) diffusion model significantly, without compromising the end performance. In a broader comparison against SOTA models our approach is substantially more efficient and compares favorably in terms of image quality. We believe that this work motivates more emphasis on the prioritization of both performance and computational accessibility.
2023-05-31
ArXiv (preprint)
arxiv.org
arxiv.org
Reinforcement learning for communication load balancing: approaches and challenges
Di Wu
Jimmy Li
Amal Ferini
Yi Tian Xu
M. Jenkin
Seowoo Jang
Xue Liu
Gregory Dudek
The amount of cellular communication network traffic has increased dramatically in recent years, and this increase has led to a demand for e… (see more)nhanced network performance. Communication load balancing aims to balance the load across available network resources and thus improve the quality of service for network users. Most existing load balancing algorithms are manually designed and tuned rule-based methods where near-optimality is almost impossible to achieve. Furthermore, rule-based methods are difficult to adapt to quickly changing traffic patterns in real-world environments. Reinforcement learning (RL) algorithms, especially deep reinforcement learning algorithms, have achieved impressive successes in many application domains and offer the potential of good adaptabiity to dynamic changes in network load patterns. This survey presents a systematic overview of RL-based communication load-balancing methods and discusses related challenges and opportunities. We first provide an introduction to the load balancing problem and to RL from fundamental concepts to advanced models. Then, we review RL approaches that address emerging communication load balancing issues important to next generation networks, including 5G and beyond. Finally, we highlight important challenges, open issues, and future research directions for applying RL for communication load balancing.
2023-05-30
Frontiers of Computer Science (published)
doi.org
Responsible Design Patterns for Machine Learning Pipelines
Saud Hakem Al Harbi
Lionel Nganyewou Tidjon
Foutse Khomh
Integrating ethical practices into the AI development process for artificial intelligence (AI) is essential to ensure safe, fair, and respon… (see more)sible operation. AI ethics involves applying ethical principles to the entire life cycle of AI systems. This is essential to mitigate potential risks and harms associated with AI, such as algorithm biases. To achieve this goal, responsible design patterns (RDPs) are critical for Machine Learning (ML) pipelines to guarantee ethical and fair outcomes. In this paper, we propose a comprehensive framework incorporating RDPs into ML pipelines to mitigate risks and ensure the ethical development of AI systems. Our framework comprises new responsible AI design patterns for ML pipelines identified through a survey of AI ethics and data management experts and validated through real-world scenarios with expert feedback. The framework guides AI developers, data scientists, and policy-makers to implement ethical practices in AI development and deploy responsible AI systems in production.
2023-05-30
ArXiv (preprint)
doi.org
arxiv.org
Spotlight Attention: Robust Object-Centric Learning With a Spatial Locality Prior
Ayush K Chakravarthy
Trang M. Nguyen
Anirudh Goyal
Yoshua Bengio
Michael Curtis Mozer
2023-05-30
ArXiv (preprint)
doi.org
arxiv.org
A graphical user interface for calculating the arterial input function during dynamic positron emission tomography
Y. Daoud
Liam Carroll
S. Enger
Purpose. Dynamic positron emission tomography (dPET) requires the acquisition of the arterial input function (AIF), conventionally obtained … (see more)via invasive arterial blood sampling. To obtain the AIF non-invasively, our group developed and combined two novel solutions consisting of (1) a detector, placed on a patient’s wrist during the PET scans to measure the radiation leaving the wrist and (2) a Geant4-based Monte Carlo simulation software. The simulations require patient-specific wrist geometry. The aim of this study was to develop a graphical user interface (GUI) allowing the user to import 2D ultrasound scans of a patient’s wrist, and measure the wrist features needed to calculate the AIF. Methods. The GUI elements were implemented using Qt5 and VTK-8.2.0. The user imports a patient’s wrist ultrasound scans, measures the radial artery and veins’ surface and depth to model a wrist phantom, then specifies the radioactive source used during the dPET scan. The phantom, the source, and the number of decay events are imported into the Geant4-based Monte Carlo software to run a simulation. In this study, 100 million decays of 18F and 68Ga were simulated in a wrist phantom designed based on an ultrasound scan. The detector’s efficiency was calculated and the results were analyzed using a clinical data processing algorithm developed in a previous study. Results. The detector’s total efficiency decreased by 3.5% for 18F and by 51.7% for 68Ga when using a phantom based on ultrasound scans compared to a generic wrist phantom. Similarly, the data processing algorithm’s accuracy decreased when using the patient-specific phantom, giving errors greater than 1.0% for both radioisotopes. Conclusions. This toolkit enables the user to run Geant4-based Monte Carlo simulations for dPET detector development applications using a patient-specific wrist phantom. Leading to a more precise simulation of the developed detector during dPET and the calculation of a personalized AIF.
2023-05-29
Physics in Medicine & Biology (published)
doi.org
The neuroconnectionist research programme
Adrien Doerig
R. Sommers
Katja Seeliger
Blake Aaron Richards
J. Ismael
Grace W. Lindsay
Konrad Paul Kording
Talia Konkle
M. Gerven
Nikolaus Kriegeskorte
Tim Kietzmann
2023-05-29
Nature Reviews Neuroscience (published)
doi.org
arxiv.org
Motor Cortex Latent Dynamics Encode Spatial and Temporal Arm Movement Parameters Independently
Andrea Colins Rodriguez
Matt G. Perich
Lee E. Miller
Mark D. Humphries
The fluid movement of an arm requires multiple spatiotemporal parameters to be set independently. Recent studies have argued that arm moveme… (see more)nts are generated by the collective dynamics of neurons in motor cortex. An untested prediction of this hypothesis is that independent parameters of movement must map to independent components of the neural dynamics. Using a task where three male monkeys made a sequence of reaching movements to randomly placed targets, we show that the spatial and temporal parameters of arm movements are independently encoded in the low-dimensional trajectories of population activity in motor cortex: each movement’s direction corresponds to a fixed neural trajectory through neural state space and its speed to how quickly that trajectory is traversed. Recurrent neural network models show that this coding allows independent control over the spatial and temporal parameters of movement by separate network parameters. Our results support a key prediction of the dynamical systems view of motor cortex, and also argue that not all parameters of movement are defined by different trajectories of population activity.
2023-05-25
BioRxiv (preprint)
doi.org
Testing Feedforward Neural Networks Training Programs
Houssem Ben Braiek
Foutse Khomh
2023-05-25
ACM Transactions on Software Engineering and Methodology (published)
doi.org
arxiv.org
Video Game Bad Smells: What They Are and How Developers Perceive Them
Vittoria Nardone
Biruk Asmare Muse
Mouna Abidi
Foutse Khomh
Massimiliano Di Penta
Video games represent a substantial and increasing share of the software market. However, their development is particularly challenging as i… (see more)t requires multi-faceted knowledge, which is not consolidated in computer science education yet. This article aims at defining a catalog of bad smells related to video game development. To achieve this goal, we mined discussions on general-purpose and video game-specific forums. After querying such a forum, we adopted an open coding strategy on a statistically significant sample of 572 discussions, stratified over different forums. As a result, we obtained a catalog of 28 bad smells, organized into five categories, covering problems related to game design and logic, physics, animation, rendering, or multiplayer. Then, we assessed the perceived relevance of such bad smells by surveying 76 game development professionals. The survey respondents agreed with the identified bad smells but also provided us with further insights about the discussed smells. Upon reporting results, we discuss bad smell examples, their consequences, as well as possible mitigation/fixing strategies and trade-offs to be pursued by developers. The catalog can be used not only as a guideline for developers and educators but also can pave the way toward better automated tool support for video game developers.
2023-05-25
ACM Transactions on Software Engineering and Methodology (published)
doi.org
Model evaluation for extreme risks
Toby Shevlane
Sebastian Farquhar
Ben Garfinkel
Mary Phuong
Jess Whittlestone
Jade Leung
Daniel Kokotajlo
Nahema A. Marchal
Markus Anderljung
Noam Kolt
Lewis Ho
Divya Siddarth
Shahar Avin
W. Hawkins
Been Kim
Iason Gabriel
Vijay Bolina
Jack Clark
Yoshua Bengio
Paul F. Christiano … (see 1 more)
Allan Dafoe
2023-05-23
ArXiv (preprint)
doi.org
arxiv.org
De novo motor learning creates structure in neural activity space that shapes adaptation
Joanna C. Chang
Matthew G Perich
Lee Miller
Juan A. Gallego
Claudia Clopath
2023-05-23
bioRxiv (preprint)
doi.org
Realistically distributing object placements in synthetic training data improves the performance of vision-based object detection models
Setareh Dabiri
Vasileios Lioutas
Berend Zwartsenberg
Yunpeng Liu
Matthew Niedoba
Xiaoxuan Liang
Dylan Green
Justice Sefas
Jonathan Wilder Lavington
Frank N. Wood
Adam Ścibior
When training object detection models on synthetic data, it is important to make the distribution of synthetic data as close as possible to … (see more)the distribution of real data. We investigate specifically the impact of object placement distribution, keeping all other aspects of synthetic data fixed. Our experiment, training a 3D vehicle detection model in CARLA and testing on KITTI, demonstrates a substantial improvement resulting from improving the object placement distribution.
2023-05-23
ArXiv (preprint)
doi.org
arxiv.org