Collaborating Authors


Deep attentive variational inference


Figure 1: Overview of a local variational layer (left) and an attentive variational layer (right) proposed in this post. Attention blocks in the variational layer are responsible for capturing long-range statistical dependencies in the latent space of the hierarchy. Generative models are a class of machine learning models that are able to generate novel data samples such as fictional celebrity faces, digital artwork, and scenic images. Currently, the most powerful generative models are deep probabilistic models. This class of models uses deep neural networks to express statistical hypotheses about the data generation process, and combine them with latent variable models to augment the set of observed data with latent (unobserved) information in order to better characterize the procedure that generates the data of interest.

Machine Translation Evaluation with Cometinho


The European Association for Machine Translation (EAMT) conference is a venue where MT researchers, users and translators gather to discuss the latest advances in the industry. It is really interesting to go there and see what is going on in the European continent in terms of MT development and adoption. In this article, I want to share some ideas from the Best Paper Award of this year. Its title is "Searching for COMETINHO: The Little Metric That Could", from the research lab of Unbabel, a company based in Lisbon, Portugal that offers translation services using MT and human translators. You can find the online version of the paper in the ACL Anthology.

An Introduction to Amazon SageMaker


Amazon SageMaker helps data scientists and inventors to prepare, make, train, and deploy high- quality machine learning models by bringing together a broad set of capabilities purpose- erected for machine learning. Amazon SageMaker make available a set of solutions for the most common use cases that may be deployed readily with just a few clicks to make it easier to grow started. Amazon SageMaker is a completely accomplished machine learning service. Data scientists and developers may speedily and easily build and train machine learning models with SageMaker. They can straight deploy them into a production-ready hosted environment.

An Overview of Small object detection by Slicing Aided Hyper Inference (SAHI)


In surveillance applications, detecting tiny items and objects that are far away in the scene is practically very difficult. Because such things are represented by a limited number of pixels in the image, traditional detectors have a tough time detecting them. So, in this article, we will look at how current models fail to recognize objects at the far end, as well as a strategy presented by Fatih Akyon et al. to address this problem called Slicing Aided Hyper Inference (SAHI) Object detection is a task that involves bounding boxes and classifying them into categories to locate all positions of objects of interest in an input. Several ways have been proposed to accomplish this goal, ranging from traditional methodologies to deep learning-based alternatives. What are the 2 approaches to Object Detection?

How to scale machine learning inference for multi-tenant SaaS use cases


This post is co-written with Sowmya Manusani, Sr. Staff Machine Learning Engineer at Zendesk Zendesk is a SaaS company that builds support, sales, and customer engagement software for everyone, with simplicity as the foundation. It thrives on making over 170,000 companies worldwide serve their hundreds of millions of customers efficiently. The Machine Learning team at Zendcaesk is responsible for enhancing Customer Experience teams to achieve their best. By combining the power of data and people, Zendesk delivers intelligent products that make their customers more productive by automating manual work. Zendesk has been building ML products since 2015, including Answer Bot, Satisfaction Prediction, Content Cues, Suggested Macros, and many more.

AI is not smart enough to solve Meta's content-policing problems, whistleblowers say


Artificial intelligence is nowhere near good enough to address problems facing content moderation on Facebook, according to whistleblower Frances Haugen. Haugen appeared at an event in London Tuesday evening with Daniel Motaung, a former Facebook moderator who is suing the company in Kenya accusing it of human trafficking. Meta has praised the efficacy of its AI systems in the past. CEO Mark Zuckerberg told a Congressional hearing in March 2021 the company relies on AI to weed out over 95% of "hate speech content." In February this year Zuckerberg said the company wants to get its AI to a "human level" of intelligence.

Building NLP Powered Applications with Hugging Face Transformers


I recently finished the fantastic new Natural Language Processing with Transformers book written by a few guys on the Hugging Face team and was inspired to put some of my newfound knowledge to use with a little NLP-based project. While searching for some ideas I came across an excellent blog post by Tezan Sahu in which he built a Microsoft Edge extension to paraphrase text highlighted on your screen. The idea is that this creates the ultimate essay companion as it can help quickly understand text with the summaries and NER, and it can get those creative juices flowing with the paraphrased text and keyword synonyms. TL;DR: This repository contains all the code mentioned in this article. ML stuff can be found in the src folder and Chrome extension stuff is in the extension folder.

Hallucinating To Better Text Translation - AI Summary


We don't start off reading raw text, which requires fundamental knowledge and understanding about the world, as well as the advanced ability to interpret and infer descriptions and relationships. Rather, humans begin our language journey slowly, by pointing and interacting with our environment, basing our words and perceiving their meaning through the context of the physical and social world. With recent, significant advances in deep learning, "there's been an interesting development in how one might use non-text information -- for example, images, audio, or other grounding information -- to tackle practical tasks involving language" says Kim, because "when humans are performing language processing tasks, we're doing so within a grounded, situated world." The pairing of hallucinated images and text during inference, the team postulated, imitates that process, providing context for improved performance over current state-of-the-art techniques, which utilize text-only data. To do this, the team used an encoder-decoder architecture with two transformers, a type of neural network model that's suited for sequence-dependent data, like language, that can pay attention key words and semantics of a sentence. Moreover, Kim and Panda note, a technique like VALHALLA is still a black box, with the assumption that hallucinated images are providing helpful information, and the team plans to investigate what and how the model is learning in order to validate their methods.

SNIA Persistent Memory And Computational Storage Summit, Part 1


SNIA held its Persistent Memory and Computational Storage Summit, virtual this year, like last year. Let's explore some of the insights from that virtual conference from the first day. Dr. Yang Seok, VP of the Memory Solutions Lab at Samsung spoke about the company's SmartSSD. He argued that computational storage devices, which off-load processing from CPUs, may reduce energy consumption and thus provide a green computing alternative. He pointed out that data center energy usage has stayed flat at about 1% since 2010 (in 2020 its was 200-250 TWh per year) due to technology innovations.

Hosting Models with TF Serving on Docker


Training a Machine Learning (ML) model is only one step in the ML lifecycle. There's no purpose to ML if you cannot get a response from your model. You must be able to host your trained model for inference. There's a variety of hosting/deployment options that can be used for ML, with one of the most popular being TensorFlow Serving. TensorFlow Serving helps take your trained model's artifacts and host it for inference.