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

We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.

Adam: A Method for Stochastic Optimization

We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.

/pdf/imagenet-classification-with-deep-convolutional-neural-1190x9a6b9.pdf

ImageNet Classification with Deep Convolutional Neural Networks

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

A time‐delay neural network (TDNN) approach is presented to speech recognition that is characterized by two important properties: (1) Using multilayer arrangements of simple computing units, a TDNN can represent arbitrary nonlinear classification decision surfaces that are learned automatically using error back propagation. (2) The time‐delay arrangement enables the network to discover acoustic‐phonetic features and the temporal relationships between them independent of position in time and, hence, not blurred by temporal shifts in the input. The TDNNs are compared with the currently most popular technique in speech recognition, hidden Markov models (HMM). Extensive performance evaluation shows that the TDNN recognizes voiced stops extracted from varying phonetic contexts at an error rate four times lower (1.5% vs 6.3%) than the best of our HMMs. To perform this task, the TDNN “invented” well‐known acoustic‐phonetic features (e.g., F2 rise, F2 fall, vowel onset) as useful abstractions. It also developed alternate internal representations to link different acoustic realizations to the same concept. The TDNNs trained for other phonetic classes achieve similar high levels of performance. The integration of such smaller networks into large phonetic nets and propose strategies for the design of neural network based large vocabulary speech recognition systems is discussed.

Speech recognition using time‐delay neural networks

The Helmholtz Machine Through Time

To generalize to novel visual scenes with new viewpoints and new object poses, a visual system needs representations of the shapes of the parts of an object that are invariant to changes in viewpoint or pose. 3D graphics representations disentangle visual factors such as viewpoints and lighting from object structure in a natural way. It is possible to learn to invert the process that converts 3D graphics representations into 2D images, provided the 3D graphics representations are available as labels. When only the unlabeled images are available, however, learning to derender is much harder. We consider a simple model which is just a set of free floating parts. Each part has its own relation to the camera and its own triangular mesh which can be deformed to model the shape of the part. At test time, a neural network looks at a single image and extracts the shapes of the parts and their relations to the camera. Each part can be viewed as one head of a multi-headed derenderer. During training, the extracted parts are used as input to a differentiable 3D renderer and the reconstruction error is backpropagated to train the neural net. We make the learning task easier by encouraging the deformations of the part meshes to be invariant to changes in viewpoint and invariant to the changes in the relative positions of the parts that occur when the pose of an articulated body changes. Cerberus, our multi-headed derenderer, outperforms previous methods for extracting 3D parts from single images without part annotations, and it does quite well at extracting natural parts of human figures.

Cerberus: A Multi-headed Derenderer.

Products of Hidden Markov Models (PoHMMs) are an interesting class of generative models which have received little attention since their introduction. This may be in part due to their more computationally expensive gradient-based learning algorithm, and the intractability of computing the log likelihood of sequences under the model. In this paper, we demonstrate how the partition function can be estimated reliably via Annealed Importance Sampling. We perform experiments using contrastive divergence learning on rainfall data and data captured from pairs of people dancing. Our results suggest that advances in learning and evaluation for undirected graphical models and recent increases in available computing power make PoHMMs worth considering for complex time-series modeling tasks.

/pdf/products-of-hidden-markov-models-it-takes-n-1-to-tango-u4mhk0pogs.pdf

Products of Hidden Markov Models: it takes N>1 to tango

As the machine learning community tackles more complex and harder problems, the graphical models needed to solve such problems become larger and more complicated. As a result performing inference and learning exactly for such graphical models become ever more expensive, and approximate inference and learning techniques become ever more prominent. 
There are a variety of techniques for approximate inference and learning in the literature. This thesis contributes some new ideas in the products of experts (PoEs) class of models (Hinton, 2002), and the Bethe free energy approximations (Yedidia et al., 2001). 
For PoEs, our contribution is in developing new PoE models for continuous-valued domains. We developed RBMrate, a model for discretized continuous-valued data. We applied it to face recognition to demonstrate its abilities. We also developed energy-based models (EBMs)—flexible probabilistic models where the building blocks consist of energy terms computed using a feed-forward network. We show that standard square noiseless independent components analysis (ICA) (Bell and Sejnowski, 1995) can be viewed as a restricted form of EBMs. Extending this relationship with ICA, we describe sparse and over-complete representations of data where the inference process is trivial since it is simply an EBM. 
For Bethe free energy approximations, our contribution is a theory relating belief propagation and iterative scaling. We show that both belief propagation and iterative scaling updates can be derived as fixed point equations for constrained minimization of the Bethe free energy. This allows us to develop a new algorithm to directly minimize the Bethe free energy, and to apply the Bethe free energy to learning in addition to inference. We also describe improvements to the efficiency of standard learning algorithms for undirected graphical models (Jirousek and Preucil, 1995).

Geoffrey E. Hinton

Papers

Speech recognition using time‐delay neural networks

The Helmholtz Machine Through Time

Cerberus: A Multi-headed Derenderer.

Products of Hidden Markov Models: it takes N>1 to tango

Bethe free energy and contrastive divergence approximations for undirected graphical models