Early detection of breast cancer can increase treatment efficiency. Architectural Distortion (AD) is a very subtle contraction of the breast tissue and may represent the earliest sign of cancer. Since it is very likely to be unnoticed by radiologists, several approaches have been proposed over the years but none using deep learning techniques. To train a Convolutional Neural Network (CNN), which is a deep neural architecture, is necessary a huge amount of data. To overcome this problem, this paper proposes a data augmentation approach applied to clinical image dataset to properly train a CNN. Results using receiver operating characteristic analysis showed that with a very limited dataset we could train a CNN to detect AD in digital mammography with area under the curve (AUC = 0.74).

Archaeologists are in dire need of automated object reconstruction methods. Fragments reassembly is close to puzzle problems, which may be solved by computer vision algorithms. As they are often beaten on most image related tasks by deep learning algorithms, we study a classification method that can solve jigsaw puzzles. In this paper, we focus on classifying the relative position: given a couple of fragments, we compute their local relation (e.g. on top). We propose several enhancements over the state of the art in this domain, which is outperformed by our method by 25\%. We propose an original dataset composed of pictures from the Metropolitan Museum of Art. We propose a greedy reconstruction method based on the predicted relative positions.

We point out important problems with the common practice of using the best single model performance for comparing deep learning architectures, and we propose a method that corrects these flaws. Each time a model is trained, one gets a different result due to random factors in the training process, which include random parameter initialization and random data shuffling. Reporting the best single model performance does not appropriately address this stochasticity. We propose a normalized expected best-out-of-$n$ performance ($\text{Boo}_n$) as a way to correct these problems.

Linear encoding of sparse vectors is widely popular, but is most commonly data-independent -- missing any possible extra (but a-priori unknown) structure beyond sparsity. In this paper we present a new method to learn linear encoders that adapt to data, while still performing well with the widely used $\ell_1$ decoder. The convex $\ell_1$ decoder prevents gradient propagation as needed in standard autoencoder training. Our method is based on the insight that unfolding the convex decoder into $T$ projected gradient steps can address this issue. Our method can be seen as a data-driven way to learn a compressed sensing matrix. Our experiments show that there is indeed additional structure beyond sparsity in several real datasets. Our autoencoder is able to discover it and exploit it to create excellent reconstructions with fewer measurements compared to the previous state of the art methods.

The web is loaded with textual content, and Natural Language Processing is a standout amongst the most vital fields in Machine Learning. But when data is huge simple Machine Learning algorithms are not able to handle it and that is when Deep Learning comes into play which based on Neural Networks. However since neural networks cannot process raw text, we have to change over them through some diverse strategies of word embedding. This paper demonstrates those distinctive word embedding strategies implemented on an Amazon Review Dataset, which has two sentiments to be classified: Happy and Unhappy based on numerous customer reviews. Moreover we demonstrate the distinction in accuracy with a discourse about which word embedding to apply when.

Recurrent neural networks (RNNs) are powerful architectures to model sequential data, due to their capability to learn short and long-term dependencies between the basic elements of a sequence. Nonetheless, popular tasks such as speech or images recognition, involve multi-dimensional input features that are characterized by strong internal dependencies between the dimensions of the input vector. We propose a novel quaternion recurrent neural network (QRNN) that takes into account both the external relations and these internal structural dependencies with the quaternion algebra. Similarly to capsules, quaternions allow the QRNN to code internal dependencies by composing and processing multidimensional features as single entities, while the recurrent operation reveals correlations between the elements composing the sequence. We show that the QRNN achieves better performances in both a synthetic memory copy task and in realistic applications of automatic speech recognition. Finally, we show that the QRNN reduces by a factor of 3x the number of free parameters needed, compared to RNNs to reach better results, leading to a more compact representation of the relevant information.

Deep learning algorithms have recently gained attention due to their inherent capabilities and the application opportunities that they provide. Two of the main reasons for the success of deep learning methods are the availability of processing power and big data. Both of these two are expensive and rare commodities that present limitations to the usage and implementation of deep learning. Decentralization of the processing and data is one of the most prevalent solutions for these issues. This paper proposes a cooperative decentralized deep learning architecture. The contributors can train deep learning models with private data and share them to the cooperative data-driven applications initiated elsewhere. Shared models are fused together to obtain a better model. In this work, the contributors can both design their own models or train the models provided by the initiator. In order to utilize an efficient decentralized learning algorithm, blockchain technology is incorporated as a method of creating an incentive-compatible market. In the proposed method, Ethereum blockchain's scripting capabilities are employed to devise a decentralized deep learning mechanism, which provides much higher, collective processing power and grants access to large amounts of data, which would be otherwise inaccessible. The technical description of the mechanism is described and the simulation results are presented.

We present an instance segmentation algorithm trained and applied to a CCTV recording of beef cattle during a winter finishing period. A fully convolutional network was transformed into an instance segmentation network that learns to label each instance of an animal separately. We introduce a conceptually simple framework that the network uses to output a single prediction for every animal. These results are a contribution towards behaviour analysis in winter finishing beef cattle for early detection of animal welfare-related problems.

While accelerators such as GPUs have limited memory, deep neural networks are becoming larger and will not fit with the memory limitation of accelerators for training. We propose an approach to tackle this problem by rewriting the computational graph of a neural network, in which swap-out and swap-in operations are inserted to temporarily store intermediate results on CPU memory. In particular, we first revise the concept of a computational graph by defining a concrete semantics for variables in a graph. We then formally show how to derive swap-out and swap-in operations from an existing graph and present rules to optimize the graph. To realize our approach, we developed a module in TensorFlow, named TFLMS. TFLMS is published as a pull request in the TensorFlow repository for contributing to the TensorFlow community. With TFLMS, we were able to train ResNet-50 and 3DUnet with 4.7x and 2x larger batch size, respectively. In particular, we were able to train 3DUNet using images of size of $192^3$ for image segmentation, which, without TFLMS, had been done only by dividing the images to smaller images, which affects the accuracy.

Stochastic video prediction is usually framed as an extrapolation problem where the goal is to sample a sequence of consecutive future image frames conditioned on a sequence of observed past frames. For the most part, algorithms for this task generate future video frames sequentially in an autoregressive fashion, which is slow and requires the input and output to be consecutive. We introduce a model that overcomes these drawbacks -- it learns to generate a global latent representation from an arbitrary set of frames within a video. This representation can then be used to simultaneously and efficiently sample any number of temporally consistent frames at arbitrary time-points in the video. We apply our model to synthetic video prediction tasks and achieve results that are comparable to state-of-the-art video prediction models. In addition, we demonstrate the flexibility of our model by applying it to 3D scene reconstruction where we condition on location instead of time. To the best of our knowledge, our model is the first to provide flexible and coherent prediction on stochastic video datasets, as well as consistent 3D scene samples. Please check the project website https://bit.ly/2jX7Vyu to view scene reconstructions and videos produced by our model.