Tomas Pfister
@tomaspfister
Followers
270
Following
27
Media
4
Statuses
26
Head of AI Research @GoogleCloud, Researcher #ML #AI #computervision
Joined January 2011
Great work from the Cloud AI Research team!
Today on the blog, read all about two new frameworks that address challenges with anomaly detection — the task of distinguishing anomalous from normal data — in both unsupervised and semi-supervised settings, with state-of-the-art results in both → https://t.co/pcfXO0VjTV
0
0
2
Today on the blog, read all about two new frameworks that address challenges with anomaly detection — the task of distinguishing anomalous from normal data — in both unsupervised and semi-supervised settings, with state-of-the-art results in both → https://t.co/pcfXO0VjTV
11
53
212
Read about FormNet, a sequence model for form-based document understanding that can process the more complex layouts frequently found in form documents and achieves state-of-the-art performance using less pre-training data than conventional methods.
research.google
Posted by Chen-Yu Lee and Chun-Liang Li, Research Scientists, Google Research, Cloud AI Team Form-based document understanding is a growing researc...
3
43
172
Introducing Learning to Prompt (L2P), an #ML model training method that uses learnable task-relevant prompts to guide pre-trained models through training on sequential tasks and results in high performance in the #ContinualLearning setting. Read more → https://t.co/pIAN0ORCEq
9
88
324
Presenting a novel approach for pre-training video understanding models on untrimmed videos that leverages the teacher-student framework to convert noisy, weak labels to more effective pseudo-labels, resulting in state-of-the-art performance. Learn more ↓
research.google
Posted by Zizhao Zhang and Guanhang Wu, Software Engineers, Google Research, Cloud AI Team Video recognition is a core task in computer vision with...
3
45
131
Our recent work: a new design of Vision Transformer (ViT) by simply nesting stacked transformer layers on local regions of images via the proposed aggregation function.
The Visual Transformer has helped advance many core computer vision applications, e.g., image classification, but training can be inefficient and models lack interpretable designs. Learn how the Nested Hierarchical Transformer addresses these challenges → https://t.co/JGYUJzW7BL
0
1
6
Introducing a novel approach for interpretable, robust, and reliable deep neural networks (DNNs) that employs controllable rule representations, which do not require retraining to adjust the rule strength at inference. Learn more below ↓
9
112
479
Our recent work: Temporal Fusion Transformer (TFT), for interpretable time series forecasting. TFT has been used to help retail and logistics companies for accurate and interpretable demand forecasting, and for applications related to climate change.
Announcing the Temporal Fusion Transformer, designed specifically to handle the heterogeneity of data in multi-horizon forecasting, which achieves more accurate forecasts with increased interpretability. Read more, including real-world applications ↓
0
0
5
Announcing the Temporal Fusion Transformer, designed specifically to handle the heterogeneity of data in multi-horizon forecasting, which achieves more accurate forecasts with increased interpretability. Read more, including real-world applications ↓
research.google
Posted by Sercan O. Arik, Research Scientist and Tomas Pfister, Engineering Manager, Google Cloud Multi-horizon forecasting, i.e. predicting variab...
11
116
407
DeepCTRL will be presented at #NeuIPS2021! - We propose a novel training method that integrates rules into deep learning. - The key aspect of DeepCTRL is the user can adjust rule strength without requiring retraining-- at inference.
1
3
5
In many AI applications, it is important to learn from “rules” beyond “data”. In our recent NeurIPS paper, we propose DeepCTRL, a novel method to integrate rules into deep learning, in a way that their effect is controllable at inference. Paper link: https://t.co/ogFVG2SnHj
2
0
0
This Google AI Blog post summarizes our research from ICLR 2021 & CVPR 2021 on anomaly detection at Google Cloud. https://t.co/mJJBPstD2M
0
0
9
Today on the blog we present a 2-stage framework for anomaly detection that combines recent progress on deep representation learning and classic one-class algorithms, is simple to train, and results in state-of-the-art performance. Learn more ↓
research.google
Posted by Chun-Liang Li and Kihyuk Sohn, Research Scientists, Google Cloud Anomaly detection (sometimes called outlier detection or out-of-distribu...
4
114
387
Learn more about a new ML-based framework for epidemiology that we applied to COVID-19, including forecasts that are released to the public daily. Read all about how it was developed and has been used by large organizations ↓
research.google
Posted by Joel Shor, Software Engineer, Google Research and Sercan Arik, Research Scientist, Google Research, Cloud AI Team Over the past 20 months...
1
51
164
Excited to see our latest COVID-19 forecasting paper (AI-augmented forecasting model) from @GoogleCloud appear in Nature Digital Medicine & Google AI Blog! Used in US & Japan for creating COVID-19 testing targets, allocating resources+simulating policies. https://t.co/DB8J5tlv26
0
3
8
"Fast Sample Reweighting" is a new paper from our research group @GoogleCloud that allows you to re-weight training samples effectively without the need for additional unbiased reward data. https://t.co/sjc4bZPQ3D PS: We’re hiring! @GoogleAI @googlecloud #ML #research #ICCV2021
1
8
52
Curious about the impact of individual data samples on your #ML model and want to improve performance focusing on more valuable training data? A new approach uses #ReinforcementLearning to estimate the value of individual data samples. Learn more ↓ https://t.co/5sLfDz4NYf
6
190
701
A nice post describing our ICML’20 paper on quantifying the value of training data using a novel approach based on meta-learning, showing which training samples are important and can be used to improve performance by removing less important samples. https://t.co/Uz8hkjMDzq
research.google
Posted by Jinsung Yoon and Sercan O. Arik, Research Scientists, Cloud AI Team, Google Research Recent work suggests that not all data samples are e...
0
0
1
In partnership with @HarvardGH, @GoogleCloud is releasing the COVID-19 Public Forecasts to help first responders and public officials track and predict future cases. Learn more about the #COVID19 Public Forecasts here:
2
41
87
After 5 months of hard work, in partnership with Harvard Global Health Institute we are pleased to release the COVID-19 Public Forecasts to help first responders and public officials track and predict future cases. Huge thank you to everyone who made this possible!
Our new COVID-19 Public Forecasts, created in partnership with @HarvardGH, are intended to serve as a resource for first responders in health care, the public sector and other organizations preparing for what lies ahead.
0
0
1