In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Very Deep Convolutional Networks for Large-Scale Image Recognition

We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames. Finally, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models. Deep Learning can be used by undergraduate or graduate students planning careers in either industry or research, and by software engineers who want to begin using deep learning in their products or platforms. A website offers supplementary material for both readers and instructors.

Deep Learning

Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters.

/pdf/batch-normalization-accelerating-deep-network-training-by-3dkqye6twt.pdf

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Web quotes are gathered from web pages that link to a web page of interest. The web quote may include text from the paragraphs that contain the hypertext links to the page of interest as well as text from other portions of the linked web page, such as text from a nearby header. The obtained web quotes may be ranked based on quality or relevance and may then be incorporated into a search engine's document index or into summary information returned to users in response to a search query.

Using text surrounding hypertext links when indexing and generating page summaries

Financial Deregulation and Offshore Banking: Lessons for Malta from Australasian/Asia-Pacific Experience

Over the past five years, deep learning and large-scale neural networks have made significant advances in speech recognition, computer vision, language understanding and translation, robotics, and many other fields. Deep learning allows the use of very raw forms of data in order to build higher-level understanding of data automatically, and can also be used to learn to accomplish complex tasks. In the next decade, it is likely that a fruitful direction for research in data management will be in how to seamlessly integrate these kinds of machine learning models into systems that store and manage data. In this talk, I will highlight some of the advances that have been made in deep learning and suggest some interesting directions for future research.

/pdf/building-machine-learning-systems-that-understand-48az1i1kbw.pdf

Building Machine Learning Systems that Understand

A method of accessing data includes storing a table that includes a plurality of tablets corresponding to distinct non-overlapping table portions. Respective pluralities of tablet access objects and application objects are stored in a plurality of servers. A distinct application object and distinct tablet are associated with each tablet access object. Each application object corresponds to a distinct instantiation of an application associated with the table. The tablet access objects and associated application objects are redistributed among the servers in accordance with a first load-balancing criterion. A first request directed to a respective tablet is received from a client. In response, the tablet access object associated with the respective tablet is used to perform a data access operation on the respective tablet, and the application object associated with the respective tablet is used to perform an additional computational operation to produce a result to be returned to the client.

Associating Application-Specific Methods with Tables Used for Data Storage

PROBLEM TO BE SOLVED: To provide a device for sampling many potentially simultaneous instruction in a processor pipeline. SOLUTION: This device is the one which samples many simultaneous execution instructions in a processor pipeline of a system 100 and a pipeline 111 has plural processing steps. This device identifies the plural instructions to be selected when the instructions are fetched to the pipeline 111. Plural selected subsets of the instructions are simultaneously executed in the pipeline 111. While any of the plural selected instructions is at any step of the pipeline 111, the state information of the system is sampled. When all the selected instructions go out of the pipeline, software is notified of it and therefore, the software can read any state information.

Jeffrey Dean

Papers

Using text surrounding hypertext links when indexing and generating page summaries

Financial Deregulation and Offshore Banking: Lessons for Malta from Australasian/Asia-Pacific Experience

Building Machine Learning Systems that Understand

Associating Application-Specific Methods with Tables Used for Data Storage

Device for sampling many potentially simultaneous instructions in processor pipeline