A5042411312- Protein Structure and Dynamics
- Computational Drug Discovery Methods
- Visual Attention and Saliency Detection
- Receptor Mechanisms and Signaling
- Aesthetic Perception and Analysis
- Image and Video Quality Assessment
- Gaze Tracking and Assistive Technology
- Molecular spectroscopy and chirality
- Scientific Computing and Data Management
- Cryptography and Data Security
- Robot Manipulation and Learning
- Spectroscopy and Quantum Chemical Studies
- Mass Spectrometry Techniques and Applications
- Bioinformatics and Genomic Networks
- Pharmacological Effects and Assays
- Stochastic Gradient Optimization Techniques
- Neurobiology and Insect Physiology Research
- Alzheimer's disease research and treatments
- Privacy-Preserving Technologies in Data
- Video Analysis and Summarization
- Multimodal Machine Learning Applications
- Robotics and Automated Systems
- Brain Tumor Detection and Classification
- Mind wandering and attention
- 3D Shape Modeling and Analysis
Google (United States)
2012-2024
Stanford University
2011-2013
University of Cambridge
2009-2011
San Francisco State University
2011
Constructor University
2006
We introduce tensor field neural networks, which are locally equivariant to 3D rotations, translations, and permutations of points at every layer. rotation equivariance removes the need for data augmentation identify features in arbitrary orientations. Our network uses filters built from spherical harmonics; due mathematical consequences this filter choice, each layer accepts as input (and guarantees output) scalars, vectors, higher-order tensors, geometric sense these terms. demonstrate...
This paper presents the Dexterity Network (Dex-Net) 1.0, a dataset of 3D object models and sampling-based planning algorithm to explore how Cloud Robotics can be used for robust grasp planning. The uses Multi- Armed Bandit model with correlated rewards leverage prior grasps in growing that currently includes over 10,000 unique 2.5 million parallel-jaw grasps. Each an estimate probability force closure under uncertainty gripper pose friction. Dex-Net 1.0 Multi-View Convolutional Neural...
Eye tracking has been widely used for decades in vision research, language and usability. However, most prior research focused on large desktop displays using specialized eye trackers that are expensive cannot scale. Little is known about movement behavior phones, despite their pervasiveness amount of time spent. We leverage machine learning to demonstrate accurate smartphone-based without any additional hardware. show the accuracy our method comparable state-of-the-art mobile 100x more...
We present a method, CamShift, for the rapid and accurate prediction of NMR chemical shifts from protein structures. The calculations performed by CamShift are based on an approximate expression in terms polynomial functions interatomic distances. Since these very fast to compute readily differentiable, approach can be utilized standard structure calculation protocols.
The stabilization of native states proteins is a powerful drug discovery strategy. It still unclear, however, whether this approach can be applied to intrinsically disordered proteins. Here, we report small molecule that stabilizes the state Aβ42 peptide, an protein fragment associated with Alzheimer's disease. We show takes place by binding mechanism, in which both and peptide remain disordered. This mechanism involves enthalpically favorable local π-stacking interactions coupled...
Behavioural assays represent sensitive methods for detecting neuronal dysfunction in model organisms. A number of manual have been established Drosophila, however these are time-consuming and generate parameter-poor phenotype descriptors. Here, we developed an automated computer vision system to monitor accurately the three-dimensional locomotor trajectories flies. This approach allows quantitative description fly trajectories, using small cohorts short acquisition times. The application...
Using only a model that was trained to predict where people look at images, and no additional training data, we can produce range of powerful editing effects for reducing distraction in images. Given an image mask specifying the region edit, backpropagate through state-of-the-art saliency parameterize differentiable operator, such within masked is reduced. We demonstrate several operators, including: recoloring which learns apply color transform camouflages blends distractors into their...
Journal Article The iFly tracking system for an automated locomotor and behavioural analysis of Drosophila melanogaster Get access Kai J. Kohlhoff, Kohlhoff Department Chemistry, University Cambridge, UK Search other works by this author on: Oxford Academic Google Scholar Thomas R. Jahn, Jahn UKDepartment Genetics, David A. Lomas, Lomas Medicine, UKCambridge Institute Medical Research, Christopher M. Dobson, Dobson Damian C. Crowther, Crowther Michele Vendruscolo E-mail: mv245@cam.ac.uk...
We present an implementation of parallel K-means clustering, called Kps-means, that achieves high performance with near-full occupancy compute kernels without imposing limits on the number dimensions and data points permitted as input, thus combining flexibility degrees parallelism efficiency. As a key element to improvement, we introduce sorting preprocessing updating steps. Our final for Nvidia GPUs speedups up 200-fold over CPU reference code three orders magnitude when compared popular...
Scientific discovery is transitioning from a focus on data collection to an emphasis analysis and prediction using large-scale computation. With appropriate software support, scientists can do these computations with unused cycles in commercial clouds. Moving science into the cloud will promote sharing collaborations that accelerate scientific discovery.
Abstract Motivation: Data clustering techniques are an essential component of a good data analysis toolbox. Many current bioinformatics applications inherently compute-intense and work with very large datasets. Sequential algorithms inadequate for providing the necessary performance. For this reason, we have created Clustering Algorithms Massively Parallel Architectures, Including GPU Nodes (CAMPAIGN), central resource tools that implemented specifically execution on massively parallel...
We consider the task of producing heatmaps from users' aggregated data while protecting their privacy. give a differentially private (DP) algorithm for this and demonstrate its advantages over previous algorithms on real-world datasets. Our core algorithmic primitive is DP procedure that takes in set distributions produces an output close Earth Mover's Distance (EMD) to average inputs. prove theoretical bounds error our under certain sparsity assumption these are essentially optimal.
Everyone is unique. Given the same visual stimuli, people's attention driven by both salient cues and their own inherent preferences. Knowledge of preferences not only facilitates understanding fine-grained patterns diverse users, but also has potential benefiting development customized applications. Nevertheless, existing saliency models typically limit scope to as it applies general population ignore variability between users' behaviors. In this paper, we identify critical roles in...
Abstract Atypical gaze patterns are a promising biomarker of autism spectrum disorder. To measure accurately, however, it typically requires highly controlled studies in the laboratory using specialized equipment that is often expensive, thereby limiting scalability these approaches. Here we test whether recently developed smartphone‐based estimation method could overcome such limitations and take advantage ubiquity smartphones. As proof‐of‐principle, measured while small sample...
Early detection of age-related diseases will greatly benefit from a model the underlying biological aging process. In this paper, we develop brain-age predictor by using structural magnetic resonance imaging (SMRI) and deep learning evaluate predicted brain age as marker brain-aging in Alzheimer's disease. Our approach does not require any domain knowledge that it trains end-to-end on SMRI image itself, has been validated real data collected elderly subjects. We developed two different...
Recent Text-to-Image (T2I) generation models such as Stable Diffusion and Imagen have made significant progress in generating high-resolution images based on text descriptions. However, many generated still suffer from issues artifacts/implausibility, misalignment with descriptions, low aesthetic quality. Inspired by the success of Reinforcement Learning Human Feedback (RLHF) for large language models, prior works collected human-provided scores feedback trained a reward model to improve T2I...
Abstract The discovery that disordered proteins are widespread in the human proteome has prompted quest for methods to characterize conformational properties determine their functional and dysfunctional behaviour. It become customary describe these terms of structural ensembles free energy landscapes, which offer thermodynamic insight. However, a current major challenge is generalize this description ‘kinetic ensembles’, thereby also providing information on transition rates between states....