AI DIGEST #MARCH20

By Alisher Abdulkhaev and Suzana Ilić

Issue #9: March 2020

  • Keras turned 5
  • Facebook releases SynSin
  • Turing Award 2019 winners: Patrick M. Hanrahan and Edwin E. Catmull
  • NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
  • Stanford NLP Group releases Stanza
  • Google announced TensorFlow Developer Certificate
  • ICLR 2020 as a fully virtual conference due to the coronavirus spreading
  • Google announced TensorFlow Quantum: An Open Source Library for Quantum Machine Learning
  • Researchers from Microsoft have created AI ethics checklist with nearly 50 engineers from a dozen tech companies
  • Google announced a new neural-network library: Neural Tangents

Keras turned 5

Happy Birthday Keras! 🎉 Creator François Chollet tweeted about it: Keras turned 5 years old!

Keras is a high-level neural networks API, written in Python and capable of running on top of TensorFlow, CNTK, or Theano. It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research.

  • Allows for easy and fast prototyping
  • Supports both convolutional networks and recurrent networks, as well as combinations of the two
  • Runs seamlessly on CPU and GPU

Documentation at Keras.io.


Facebook releases SynSin

SynSin: End-to-end View Synthesis from a Single Image (CVPR 2020)

Facebook released the code for SynSin, it allows for synthesising of new views of a scene given a single image of an unseen scene at test time. It is trained with pairs of views in a self-supervised fashion, end to end, using GAN techniques and a new differentiable point cloud renderer. At test time, a single image of an unseen scene is input to the model from which new views are generated.

Paper | GitHub


Turing Award 2019 winners: Patrick M. Hanrahan and Edwin E. Catmull

Patrick M. Hanrahan and Edwin E. Catmull win Turing Award 2019 – often described as the “Nobel Prize” of computing – for fundamental contributions to 3D computer graphics, and the impact of computer-generated imagery (CGI) in filmmaking and other applications. Patrick M. Hanrahan is a Canon Professor at Stanford. Edwin E. Catmull is an American retired computer scientist and former president of Pixar and Walt Disney Animation Studios.

Source: Stanford NewsACM

Left: Patrick M. Hanrahan | Right: Edwin E. Catmull


NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis

Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng

Paper | Website

– From the orignal abstract:

NeRF achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. The algorithm represents a scene using a fully-connected (non-convolutional) deep network, whose input is a single continuous 5D coordinate and whose output is the volume density and view-dependent emitted radiance at that spatial location. Views are synthesized by querying 5D coordinates along camera rays and classic volume rendering techniques are used to project the output colors and densities into an image. Because volume rendering is naturally differentiable, the only input required to optimize the representation is a set of images with known camera poses. The researchers describe how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrate results that outperform prior work on neural rendering and view synthesis.


Stanford NLP Group releases Stanza

Stanford NLP announced Stanza v1.0.0, the new packaging of the Python NLProc library for many human languages (now including mainland Chinese, Putonghua written with simplified characters), greatly improved and including NER.

Features

  • Native Python implementation requiring minimal efforts to set up;
  • Full neural network pipeline for robust text analytics, including tokenization, multi-word token (MWT) expansion, lemmatization, part-of-speech (POS) and morphological features tagging, dependency parsing, and named entity recognition;
  • Pretrained neural models supporting 66 (human) languages;
  • A stable, officially maintained Python interface to CoreNLP.

References


Google announced TensorFlow Developer Certificate

  • helps developers ML to showcase their ML proficiency
  • helps companies hire ML developers to solve challenging problems
  • this certificate program requires an understanding of building basic TensorFlow models using Computer Vision, Sequence modeling, and Natural Language Processing
  • by passing the exam you will receive an official TensorFlow Developer Certificate and badge to showcase your skill set and can share on LinkedIn or other social networks
  • Source: TensorFlow Blog

Google announced a new neural-network library: Neural Tangents

  • easy-to-use, open-source neural network library
  • provides researchers with an opportunity of building an (in)finite widths versions of neural networks simultaneously
  • Source: Google AI Blog

AI Ethics Checklist

Researchers from Microsoft have created an AI ethics checklist with nearly 50 engineers from a dozen tech companies

  • Madaio, Michael A., Luke Stark, Jennifer Wortman Vaughan, and Hanna Wallach. “Co-Designing Checklists to Understand Organizational Challenges and Opportunities around Fairness in AI.”

From the original abstract:
Many organizations have published principles intended to guide the ethical development and deployment of AI systems; however, their abstract nature makes them difficult to operationalize. Some organizations have therefore produced AI ethics checklists, as well as checklists for more specific concepts, such as fairness, as applied to AI systems. But unless checklists are grounded in practitioners’ needs, they may be misused. To understand the role of checklists in AI ethics, we conducted an iterative co-design process with 48 practitioners, focusing on fairness. We co-designed an AI fairness checklist and identified desiderata and concerns for AI fairness checklists in general. We found that AI fairness checklists could provide organizational infrastructure for formalizing ad-hoc processes and empowering individual advocates. We discuss aspects of organizational culture that may impact the efficacy of such checklists, and highlight future research directions.


ICLR 2020 as a fully virtual conference due to the coronavirus spreading

Due to growing concerns about COVID-19, ICLR2020 will cancel its physical conference this year, instead shifting to a fully virtual conference. We were very excited to hold ICLR in Addis Ababa, and it is disappointing that we will not all be able to come together in person in April. This unfortunate event does give us the opportunity to innovate on how to host an effective remote conference. The organizing committees are now working to create a virtual conference that will be valuable and engaging for both presenters and attendees.

Source: ICLR 2020


Google announced TensorFlow Quantum: An Open Source Library for Quantum Machine Learning

Google announced the release of TensorFlow Quantum (TFQ), in collaboration with University of WaterlooX, and Volkswagen an open-source library for the rapid prototyping of quantum ML models.

  • TFQ provides the tools to bring the quantum computing and machine learning research communities together
    • to control and and model natural and artificial quantum systems
    • for example, Noisy Intermediate Scale Quantum (NISQ) processors with ~50-100 qubits
  • TFQ integrates Cirq and TensorFlow:
    • offers high-level abstractions for the design and implementation of both discriminative generative quantum-classical models
    • offers high-level abstractions for the design and implementation of both discriminative generative quantum-classical models
  • Source: Google AI Blog

Leave a Reply