Henry Tan

The computer gaming industry is large and growing rapidly. Consumers demand realistic motion in computer games. Consequently, the ability to realistically simulate motion is vital in computer games development. Currently, most simulation of motion is done through the use of a physics engine which essentially involves numerically solution of the set of differential equations which describe the corresponding physics situation in the real world. This paper proposes an alternative strategy using a… Show more

The computer gaming industry is large and growing rapidly. Consumers demand realistic motion in computer games. Consequently, the ability to realistically simulate motion is vital in computer games development. Currently, most simulation of motion is done through the use of a physics engine which essentially involves numerically solution of the set of differential equations which describe the corresponding physics situation in the real world. This paper proposes an alternative strategy using a multilayer perceptron to generate these simulations. Benchmarking against traditional physics engines shows that there are considerable advantages to the new methodology. Show less

Video Intelligence API has pre-trained machine learning models that automatically recognize a vast number of objects, places, and actions in stored and streaming video. It’s highly efficient for common use cases and improves over time as new concepts are introduced.

Large deep neural network models have recently demonstrated state-of-the-art accuracy on hard visual recognition tasks. Unfortunately such models are extremely time consuming to train and require large amount of compute cycles. We describe the design and implementation of a distributed system called Adam comprised of commodity server machines to train such models that exhibits world-class performance, scaling and task accuracy on visual recognition tasks. Adam achieves high efficiency and… Show more

Large deep neural network models have recently demonstrated state-of-the-art accuracy on hard visual recognition tasks. Unfortunately such models are extremely time consuming to train and require large amount of compute cycles. We describe the design and implementation of a distributed system called Adam comprised of commodity server machines to train such models that exhibits world-class performance, scaling and task accuracy on visual recognition tasks. Adam achieves high efficiency and scalability through whole system co-design that optimizes and balances workload computation and communication. We exploit asynchrony throughout the system to improve performance and show that it additionally improves the accuracy of trained models. Adam is significantly more efficient and scalable than was previously thought possible and used 30x fewer machines to train a large 2 billion connection model to 2x higher accuracy in comparable time on the ImageNet 22,000 category image classification task than the system that previously held the record for this benchmark. We also show that task accuracy improves with larger models. Our results provide compelling evidence that a distributed systems-driven approach to deep learning using current training algorithms is worth pursuing. Show less

Association mining consists of two important problems, namely frequent patterns
discovery and rule construction. Because of its importance and application in a number of data mining
tasks, it has become the focus of many studies. A substantial amount of research has
gone into the development of efficient algorithms for mining patterns from large
structured or relational data. Compared with the fruitful achievements in mining
structured data, mining in the semi-structured… Show more

Association mining consists of two important problems, namely frequent patterns
discovery and rule construction. Because of its importance and application in a number of data mining
tasks, it has become the focus of many studies. A substantial amount of research has
gone into the development of efficient algorithms for mining patterns from large
structured or relational data. Compared with the fruitful achievements in mining
structured data, mining in the semi-structured world still remains at a preliminary stage.
The most popular representative of the semi-structured data is XML. Mining frequent
patterns from XML poses more challenges in comparison to mining frequent patterns
from relational data because XML is a tree-structured data and has an ordered data
context. Moreover, XML data in general is larger in data size due to richer contents and
more meta-data. The increase of
XML data and the need for mining semi-structured data has sparked a lot of interest in
finding frequent rooted trees in forests.

In this thesis, we aim to develop a framework to mine frequent patterns from XML
documents. The framework utilizes a structure-guided enumeration approach, Tree
Model Guided (TMG), for efficient enumeration of tree structure and it makes use of
novel structures for fast enumeration and frequency counting. By utilizing a novel
array-based structure, an embedded list (EL), the framework offers a simple sequence like tree enumeration technique. The effectiveness and extendibility of the framework is
demonstrated in that it can be utilized not only for enumerating ordered subtrees but
also for enumerating unordered subtrees and subsequences. The framework
tackles the unprecedented complexity in mining frequent tree-structured patterns by
generating only valid candidates with non-zero frequency count and employing a
constraint-driven approach. Show less

Empowering businesses with Google Cloud AI
Machine learning has produced business and research breakthroughs ranging from network security to medical diagnoses. We built the Tensor Processing Unit (TPU) in order to make it possible for anyone to achieve similar breakthroughs. Cloud TPU is the custom-designed machine learning ASIC that powers Google products like Translate, Photos, Search, Assistant, and Gmail. Here’s how you can put the TPU and machine learning to work accelerating your… Show more

Empowering businesses with Google Cloud AI
Machine learning has produced business and research breakthroughs ranging from network security to medical diagnoses. We built the Tensor Processing Unit (TPU) in order to make it possible for anyone to achieve similar breakthroughs. Cloud TPU is the custom-designed machine learning ASIC that powers Google products like Translate, Photos, Search, Assistant, and Gmail. Here’s how you can put the TPU and machine learning to work accelerating your company’s success, especially at scale. Show less

Maguro is a system for efficiently searching very large collections of text content of up to 1 trillion documents at low cost. Search engines span across content that is very dynamic and highly augmented with metadata to the tail content of the web. A long tail distribution of content calls for different trade-offs in the design space for good efficiency across the entire index range. Maguro is designed for the long tail of content with less dynamics and less metadata, but very good cost… Show more

Maguro is a system for efficiently searching very large collections of text content of up to 1 trillion documents at low cost. Search engines span across content that is very dynamic and highly augmented with metadata to the tail content of the web. A long tail distribution of content calls for different trade-offs in the design space for good efficiency across the entire index range. Maguro is designed for the long tail of content with less dynamics and less metadata, but very good cost efficiency. Maguro is part of the serving stack in Bing and allows us to scale the index significantly better. Show less

The core open source library to help you develop and train ML models. Get started quickly by running Colab notebooks directly in your browser.