Deep Learning Weekly Issue #158

Apple's commitment to on-device AI; shrinking DL's carbon footprint; style transfer for medical data; and more

Hey folks,

This week in deep learning we bring you MediaPipe Iris: Real-time Iris Tracking & Depth Estimation, ‘inconsistent’ benchmarking across 3,867 AI research papers, the Weight Clustering API from the TensorFlow Model Optimization Toolkit, and NeuralCam Live which uses ML to turn iPhones into ‘smart’ computer webcams.

You may also enjoy reading about why Apple believes it’s an AI leader, shrinking deep learning's carbon footprint, next-gen intelligent farming machines, Aligning AI With Shared Human Values and more!

As always, happy reading and hacking. If you have something you think should be in next week's issue, find us on Twitter: @dl_weekly.

Until next week!


Here’s why Apple believes it’s an AI leader—and why it says critics have it all wrong

Apple AI chief and ex-Googler John Giannandrea dives into the details with Ars.

Shrinking deep learning's carbon footprint

Through innovation in software and hardware, researchers move to reduce the financial and environmental costs of modern artificial intelligence.

Researchers find ‘inconsistent’ benchmarking across 3,867 AI research papers

Researchers analyzed over 3,000 model performance results listed on Papers with Code and found that the metrics used to benchmark AI and machine learning models often inadequately reflect those models’ true performances.

PyTorch drives next-gen intelligent farming machines

California-based Blue River Technology’s See & Spray robotic farming machine combines machine learning and computer vision to identify weeds among crops in real time and to treat weeds while leaving crops unharmed.

University of Michigan study advocates ban of facial recognition in schools

University of Michigan researchers recently published a study showing facial recognition technology in schools has limited efficacy and presents a number of serious problems.

Mobile + Edge

MediaPipe Iris: Real-time Iris Tracking & Depth Estimation

Google AI announced the release of MediaPipe Iris, a new machine learning model for accurate iris estimation.

Google adds Digital Ink Recognition API for touch and stylus input to ML Kit

Google launched the Digital Ink Recognition API on Android and iOS to allow developers to create apps where stylus and touch act as inputs.

NeuralCam Live uses ML to turn iPhones into ‘smart’ computer webcams

The premise of NeuralCam Live is to combine an iPhone’s front camera with machine learning to create a higher-quality computer video stream than a traditional webcam can deliver.

On-device Supermarket Product Recognition

Google recently released Lookout, an Android app that uses computer vision to make the physical world more accessible for users who are visually impaired.

Eta Compute introduces TENSAI Flow

Eta Compute, a company that looks to deliver machine learning to low power IoT and edge devices using its TENSAI Platform, has announced its TENSAI Flow software suite.


TensorFlow Model Optimization Toolkit — Weight Clustering API

This post covers the weight clustering technique used to build smaller and faster ML models.

Layerwise learning for Quantum Neural Networks

In this article, the author introduces a training strategy that addresses vanishing gradients in quantum neural networks (QNNs), and makes better use of the resources provided by a NISQ device.

Neural Style Transfer for Augmenting Medical Data

In this notebook, the author shows how to use neural style transfer to address covariate shift between medical image datasets.

Libraries & Code

[GitHub] linkedin/detext

DeText is a Deep Text understanding framework for NLP related ranking, classification, and language generation tasks.

[GitHub] Mawiszus/TOAD-GAN

Official repository for "TOAD-GAN: Coherent Style Level Generation from a Single Example" by Maren Awiszus, Frederik Schubert and Bodo Rosenhahn.

Papers & Publications

DeText: A Deep Text Ranking Framework with BERT

Abstract: Ranking is the most important component in a search system. Most search systems deal with large amounts of natural language data, hence an effective ranking system requires a deep understanding of text semantics. Recently, deep learning based natural language processing (deep NLP) models have generated promising results on ranking systems. BERT is one of the most successful models that learn contextual embedding, which has been applied to capture complex query-document relations for search ranking. However,this is generally done by exhaustively interacting each query word with each document word, which is inefficient for online servingin search product systems. In this paper, we investigate how to build an efficient BERT-based ranking model for industry use cases.The solution is further extended to a general ranking framework,DeText, that is open sourced and can be applied to various ranking productions. Offline and online experiments of DeText on three real-world search systems present significant improvement over state-of-the-art approaches.

Aligning AI With Shared Human Values

Abstract: We show how to assess a language model's knowledge of basic concepts of morality. We introduce the ETHICS dataset, a new benchmark that spans concepts in justice, well-being, duties, virtues, and commonsense morality. Models predict widespread moral judgments about diverse text scenarios. This requires connecting physical and social world knowledge to value judgements, a capability that may enable us to filter out needlessly inflammatory chatbot outputs or eventually regularize open-ended reinforcement learning agents. With the ETHICS dataset, we find that current language models have a promising but incomplete understanding of basic ethical knowledge. Our work shows that progress can be made on machine ethics today, and it provides a stepping stone toward AI that is aligned with human values.