Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

/pdf/deep-residual-learning-for-image-recognition-b75jg61qrq.pdf

Deep Residual Learning for Image Recognition

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.

Long short-term memory

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

Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. 
The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

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

N400 amplitudes reflect change in a probabilistic representation of meaning: Evidence from a connectionist model.

Illusory conjunctions in normal and simultanagnosic subjects are two instances where the visual features of multiple objects are incorrectly 'bound' together. A connectionist model explores how multiple objects could be perceived correctly in normal subjects given sufficient time, but could give rise to illusory conjunctions with damage or time pressure. In this model, perception of two objects benefits from lateral connections between hidden layers modelling aspects of the ventral and dorsal visual pathways. As with simultanagnosia, simulations of dorsal lesions impair multi-object recognition. In contrast, a large ventral lesion has minimal effect on dorsal functioning, akin to dissociations between simple object manipulation (retained in visual form agnosia and semantic dementia) and object discrimination (impaired in these disorders) [Hodges, J.R., Bozeat, S., Lambon Ralph, M.A., Patterson, K., and Spatt, J. (2000), 'The Role of Conceptual Knowledge: Evidence from Semantic Dementia', Brain, 123, 1913-1925; Milner, A.D., and Goodale, M.A. (2006), The Visual Brain in Action (2nd ed.), New York: Oxford]. It is hoped that the functioning of this model might suggest potential processes underlying dorsal and ventral contributions to the correct perception of multiple objects.

A PDP model of the simultaneous perception of multiple objects

Interest in the nature of conceptual knowledge extends back at least to the ancient Greek philosophers. In recent years, there has been a wide range of different approaches to understanding the nature of conceptual knowledge, its development, and its neural basis. In most other work, however, these issues are not all treated together. Instead, workers in philosophy, adult experimental psychology, child development, and cognitive neuroscience have pursued related questions in relative ignorance of each other's efforts. Even within cognitive neuroscience, there has been until recently a relative separation between approaches taken by neuropsychologists, who study the effects of brain disease on cognition in patients, and researchers who study the neural basis of conceptual knowledge in neurologically intact populations, using functional imaging and related methods.

/pdf/semantic-cognition-its-nature-its-development-and-its-neural-2q5tvzcmf9.pdf

Semantic cognition: Its nature, its development and its neural basis.

Complementary processing systems: A PDP model of the simultaneous perception of multiple objects Cynthia Henderson Stanford University James McClelland Stanford University Abstract: Illusory conjunctions in normal and simultanagnosic subjects are instances where the binding of visual information fails to function correctly. When presented with multiple objects simultaneously, simultanagnosic pa- tients and normal subjects under conditions of attentional loads or brief presentation times often erroneously report miscombinations of features of the objects. A connectionist model of multi-object perception examines how the concurrent perception of more than one object could occur in normal subjects and become deficient with shortened processing times. In this model, the correct identification of two objects is accomplished through lateral connections between the ventral and dorsal pathways. Lesioning of the dorsal pathway produces failures in multi-object recogni- tion characteristic of the effect of parietal damage in simultanagnosia. It is hoped that the functioning of this model might help elucidate possible processes underlying the correct solution of the binding problem in normal subjects.

/pdf/complementary-processing-systems-a-pdp-model-of-the-58wpd59k37.pdf

Complementary processing systems: A PDP model of the simultaneous perception of multiple objects

Traditional models of language processing process language by rule. This approach faces two p. oblems. First, there are difficulties in using the rules during processing, since often one rule must be pitted against another. In this case traditional approaches face the difficult problem of deciding which rule should win in such cases. Second, there are difficulties in acquiring rules, since it is often hard to know when a rule should be proposed, or when a sentence should be handled as one of many special cases.

James L. McClelland

Papers

N400 amplitudes reflect change in a probabilistic representation of meaning: Evidence from a connectionist model.

A PDP model of the simultaneous perception of multiple objects

Semantic cognition: Its nature, its development and its neural basis.

Complementary processing systems: A PDP model of the simultaneous perception of multiple objects

Connectionist Models of Language