Let's continue learning about misconceptions around artificial neural networks. In Part 1, we discussed the most simple neural network architecture: the multi-layer perceptron. There are many different neural network architectures (far too many to mention here) and the performance of any neural network is a function of its architecture and weights. Many modern-day advances in the field of machine learning do not come from rethinking the way that perceptrons and optimization algorithms work but rather from being creative regarding how these components fit together. Below, I discuss some very interesting and creative neural network architectures that have developed over time.

Researchers in artificial intelligence can stand to make a ton of money. But this week, we actually know just how much some A.I. experts are being paid -- and it's a lot, even at a nonprofit. OpenAI, a nonprofit research lab, paid its lead A.I. expert, Ilya Sutskever, more than $1.9 million in 2016, according to a recent public tax filing. Another researcher, Ian Goodfellow, made more than $800,000 that year, even though he was only hired in March, the New York Times reported. As the publication points out, the figures are eye-opening and offer a bit of insight on how much A.I. researchers are being paid across the globe.

Classification of human emotions remains an important and challenging task for many computer vision algorithms, especially in the era of humanoid robots which coexist with humans in their everyday life. Currently proposed methods for emotion recognition solve this task using multi-layered convolutional networks that do not explicitly infer any facial features in the classification phase. In this work, we postulate a fundamentally different approach to solve emotion recognition task that relies on incorporating facial landmarks as a part of the classification loss function. To that end, we extend a recently proposed Deep Alignment Network (DAN), that achieves state-of-the-art results in the recent facial landmark recognition challenge, with a term related to facial features. Thanks to this simple modification, our model called EmotionalDAN is able to outperform state-of-the-art emotion classification methods on two challenging benchmark dataset by up to 5%.

Deep learning approaches have shown remarkable performance in many areas of pattern recognition recently. In spite of their power in hierarchical feature extraction and classification, this type of neural network is computationally expensive and difficult to implement on hardware for portable devices. In an other vein of research on neural network architectures, spiking neural networks (SNNs) have been described as power-efficient models because of their sparse, spike-based communication framework. SNNs are brain-inspired such that they seek to mimic the accurate and efficient functionality of the brain. Recent studies try to take advantages of the both frameworks (deep learning and SNNs) to develop a deep architecture of SNNs to achieve high performance of recently proved deep networks while implementing bio-inspired, power-efficient platforms. Additionally, As the brain process different stimuli patterns through multi-layer SNNs that are communicating by spike trains via adaptive synapses, developing artificial deep SNNs can also be very helpful for understudying the computations done by biological neural circuits. Having both computational and experimental backgrounds, we are interested in including a comprehensive summary of recent advances in developing deep SNNs that may assist computer scientists interested in developing more advanced and efficient networks and help experimentalists to frame new hypotheses for neural information processing in the brain using a more realistic model.

Inner product-based convolution has been a central component of convolutional neural networks (CNNs) and the key to learning visual representations. Inspired by the observation that CNN-learned features are naturally decoupled with the norm of features corresponding to the intra-class variation and the angle corresponding to the semantic difference, we propose a generic decoupled learning framework which models the intra-class variation and semantic difference independently. Specifically, we first reparametrize the inner product to a decoupled form and then generalize it to the decoupled convolution operator which serves as the building block of our decoupled networks. We present several effective instances of the decoupled convolution operator. Each decoupled operator is well motivated and has an intuitive geometric interpretation. Based on these decoupled operators, we further propose to directly learn the operator from data. Extensive experiments show that such decoupled reparameterization renders significant performance gain with easier convergence and stronger robustness.

One of the most significant bottleneck in training large scale machine learning models on parameter server (PS) is the communication overhead, because it needs to frequently exchange the model gradients between the workers and servers during the training iterations. Gradient quantization has been proposed as an effective approach to reducing the communication volume. One key issue in gradient quantization is setting the number of bits for quantizing the gradients. Small number of bits can significantly reduce the communication overhead while hurts the gradient accuracies, and vise versa. An ideal quantization method would dynamically balance the communication overhead and model accuracy, through adjusting the number bits according to the knowledge learned from the immediate past training iterations. Existing methods, however, quantize the gradients either with fixed number of bits, or with predefined heuristic rules. In this paper we propose a novel adaptive quantization method within the framework of reinforcement learning. The method, referred to as MQGrad, formalizes the selection of quantization bits as actions in a Markov decision process (MDP) where the MDP states records the information collected from the past optimization iterations (e.g., the sequence of the loss function values). During the training iterations of a machine learning algorithm, MQGrad continuously updates the MDP state according to the changes of the loss function. Based on the information, MDP learns to select the optimal actions (number of bits) to quantize the gradients. Experimental results based on a benchmark dataset showed that MQGrad can accelerate the learning of a large scale deep neural network while keeping its prediction accuracies.

In this paper we discuss policy iteration methods for approximate solution of a finite-state discounted Markov decision problem, with a focus on feature-based aggregation methods and their connection with deep reinforcement learning schemes. We introduce features of the states of the original problem, and we formulate a smaller "aggregate" Markov decision problem, whose states relate to the features. The optimal cost function of the aggregate problem, a nonlinear function of the features, serves as an architecture for approximation in value space of the optimal cost function or the cost functions of policies of the original problem. We discuss properties and possible implementations of this type of aggregation, including a new approach to approximate policy iteration. In this approach the policy improvement operation combines feature-based aggregation with reinforcement learning based on deep neural networks, which is used to obtain the needed features. We argue that the cost function of a policy may be approximated much more accurately by the nonlinear function of the features provided by aggregation, than by the linear function of the features provided by deep reinforcement learning, thereby potentially leading to more effective policy improvement.

In this paper, we present BigDL, a distributed deep learning framework for Big Data platforms and workflows. It is implemented on top of Apache Spark, and allows users to write their deep learning applications as standard Spark programs (running directly on large-scale big data clusters in a distributed fashion). It provides an expressive, "data-analytics integrated" deep learning programming model, so that users can easily build the end-to-end analytics + AI pipelines under a unified programming paradigm; by implementing an AllReduce like operation using existing primitives in Spark (e.g., shuffle, broadcast, and in-memory data persistence), it also provides a highly efficient "parameter server" style architecture, so as to achieve highly scalable, data-parallel distributed training. Since its initial open source release, BigDL users have built many analytics and deep learning applications (e.g., object detection, sequence-to-sequence generation, visual similarity, neural recommendations, fraud detection, etc.) on Spark.

Many natural language processing tasks require dealing with Named Entities (NEs) in the texts themselves and sometimes also in external knowledge sources. While this is often easy for humans, recent neural methods that rely on learned word embeddings for NLP tasks have difficulty with it, especially with out of vocabulary or rare NEs. In this paper, we propose a new neural method for this problem, and present empirical evaluations on a structured Question-Answering task, three related Goal-Oriented dialog tasks and a reading-comprehension-based task. They show that our proposed method can be effective in dealing with both in-vocabulary and out of vocabulary (OOV) NEs. We create extended versions of dialog bAbI tasks 1,2 and 4 and Out-of-vocabulary (OOV) versions of the CBT test set which will be made publicly available online.

Distributed representation plays an important role in deep learning based natural language processing. However, the representation of a sentence often varies in different tasks, which is usually learned from scratch and suffers from the limited amounts of training data. In this paper, we claim that a good sentence representation should be invariant and can benefit the various subsequent tasks. To achieve this purpose, we propose a new scheme of information sharing for multi-task learning. More specifically, all tasks share the same sentence representation and each task can select the task-specific information from the shared sentence representation with attention mechanism. The query vector of each task's attention could be either static parameters or generated dynamically. We conduct extensive experiments on 16 different text classification tasks, which demonstrate the benefits of our architecture.