For many years, i-vector based audio embedding techniques were the dominant approach for speaker verification and speaker diarization applications. However, mirroring the rise of deep learning in various domains, neural network based audio embeddings, also known as d-vectors, have consistently demonstrated superior speaker verification performance. In this paper, we build on the success of d-vector based speaker verification systems to develop a new d-vector based approach to speaker diarization. Specifically, we combine LSTM-based d-vector audio embeddings with recent work in non-parametric clustering to obtain a state-of-the-art speaker diarization system. Our system is evaluated on three standard public datasets, suggesting that d-vector based diarization systems offer significant advantages over traditional i-vector based systems. We achieved a 12.0% diarization error rate on NIST SRE 2000 CALLHOME, while our model is trained with out-of-domain data from voice search logs.

ABSTRACT In this paper, we propose a new loss function called generalized end-to-end (GE2E) loss, which makes the training of speaker verification models more efficient than our previous tuple-based endto-end (TE2E) loss function. Unlike TE2E, the GE2E loss function updates the network in a way that emphasizes examples that are difficult to verify at each step of the training process. Additionally, the GE2E loss does not require an initial stage of example selection. We also introduce the MultiReader technique, which allows us to do domain adaptation -- training a more accurate model that supports multiple keywords (i.e., "OK Google" and "Hey Google") as well as multiple dialects. Background 1. INTRODUCTION Speaker verification (SV) is the process of verifying whether an utterance belongs to a specific speaker, based on that speaker's known utterances (i.e., enrollment utterances), with applications such as Voice Match [1, 2].

Electron Cryo-Tomography (ECT) enables 3D visualization of macromolecule structure inside single cells. Macromolecule classification approaches based on convolutional neural networks (CNN) were developed to separate millions of macromolecules captured from ECT systematically. However, given the fast accumulation of ECT data, it will soon become necessary to use CNN models to efficiently and accurately separate substantially more macromolecules at the prediction stage, which requires additional computational costs. To speed up the prediction, we compress classification models into compact neural networks with little in accuracy for deployment. Specifically, we propose to perform model compression through knowledge distillation. Firstly, a complex teacher network is trained to generate soft labels with better classification feasibility followed by training of customized student networks with simple architectures using the soft label to compress model complexity. Our tests demonstrate that our compressed models significantly reduce the number of parameters and time cost while maintaining similar classification accuracy.

But more importantly - I need to mention that Terence Parr did nearly all the work on this. He shared my passion for making something that anyone could read on any device to such an extent that he ended up creating a new tool for generating fast, mobile-friendly math-heavy texts: https://github.com/parrt/bookish .

I specialize in the analysis of time series data (a series of observations over time). I am particularly experienced in the utilities sector. I have predicted the price of energy, power, and gas with more than 98% accuracy consistently [using Mean Absolute Percentage Error (MAPE) loss function]. I can process massive streams of both unstructured and structured data almost in real time using big data analytics platforms. Recently, I was introduced to blockchain technology, and I find it fascinating!

I started this article with the hopes of confronting a few misconceptions about Deep Learning (DL), a field of Machine Learning that is simultaneously labelled a silver bullet and research hype. The truth lies somewhere in the middle, and I hope I can un-muddy the waters -- at least a little bit. Importantly, I hope to clarify some processes to attack DL problems and also discuss why it performs so well in some areas such as Natural Language Processing (NLP), image recognition, and machine-translation while failing at others. Media often portrays Deep Learning as a magical recipe to the end of the world or the solution to all life's problems. In reality, it is anything but. Moreover, while DL has its fair share of strange behaviour and unexplained results, it is ultimately meritocratically driven.

Suppose you are playing football. The ball arrives at your feet, and you decide to pass it to the unmarked striker. What seems like one simple action requires two different kinds of thought. First, you recognise that there is a football at your feet. This recognition requires intuitive perceptual thinking - you cannot easily articulate how you come to know that there is a ball at your feet, you just see that it is there.

I come from physics and computer engineering. I studied both in Venezuela, and then I did a Master in Physics in Mexico. But I consider myself a Data Scientist. So even though I have a good and extensive background in math, calculus and statistics, it was not easy to get started with machine learning and then deep learning. This subjects are not new, but the way we study them, how we build software and solutions that use them, and also the way we program or interact with them has changed dramatically.

Last year we announced that we were developing a new deep learning course based on Pytorch (and a new library we have built, called fastai), with the goal of allowing more students to be able to achieve world-class results with deep learning. Today, we are making this course, Practical Deep Learning for Coders 2018, generally available for the first time, following the completion of a preview version of the course by 600 students through our diversity fellowship, international fellowship, and Data Institute in-person programs. The only prerequisites are a year of coding experience, and high school math (math required for understanding the material is introduced as required during the course). The course includes around 15 hours of lessons and a number of interactive notebooks, and is now available for free (with no ads) at course.fast.ai. Our research focuses on how to make practically useful deep learning more widely accessible. Often we've found that the current state of the art (SoTA) approaches aren't good enough to be used in practice, so we have to figure out how to improve them.

Pinterest is very committed to improving its search technology through AI -- so committed, in fact, that it just hired one of the foremost experts in the field. The social network has announced that it's recruiting Chuck Rosenberg, Google's AI vision research leader, to become its Head of Computer Vision. He spent just shy of 14 years at Google and was responsible for a number of major AI-related efforts, including the first large deployment of an image-focused deep learning network. The exact nature of what Rosenberg is doing is under wraps, to no one's surprise, but he's expected to guide engineers as they craft "large-scale" object detection algorithms. Before Google, he worked at HP Labs and was one of iRobot's earliest employees.