We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. 
BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and moves. Here we introduce a new approach to computer Go that uses ‘value networks’ to evaluate board positions and ‘policy networks’ to select moves. These deep neural networks are trained by a novel combination of supervised learning from human expert games, and reinforcement learning from games of self-play. Without any lookahead search, the neural networks play Go at the level of stateof-the-art Monte Carlo tree search programs that simulate thousands of random games of self-play. We also introduce a new search algorithm that combines Monte Carlo simulation with value and policy networks. Using this search algorithm, our program AlphaGo achieved a 99.8% winning rate against other Go programs, and defeated the human European Go champion by 5 games to 0. This is the first time that a computer program has defeated a human professional player in the full-sized game of Go, a feat previously thought to be at least a decade away.

/pdf/mastering-the-game-of-go-with-deep-neural-networks-and-tree-3viuxo6uxm.pdf

Mastering the game of Go with deep neural networks and tree search

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

/pdf/language-models-are-few-shot-learners-2fg8gvia7m.pdf

Language Models are Few-Shot Learners

A long-standing goal of artificial intelligence is an algorithm that learns, tabula rasa, superhuman proficiency in challenging domains. Recently, AlphaGo became the first program to defeat a world champion in the game of Go. The tree search in AlphaGo evaluated positions and selected moves using deep neural networks. These neural networks were trained by supervised learning from human expert moves, and by reinforcement learning from self-play. Here we introduce an algorithm based solely on reinforcement learning, without human data, guidance or domain knowledge beyond game rules. AlphaGo becomes its own teacher: a neural network is trained to predict AlphaGo’s own move selections and also the winner of AlphaGo’s games. This neural network improves the strength of the tree search, resulting in higher quality move selection and stronger self-play in the next iteration. Starting tabula rasa, our new program AlphaGo Zero achieved superhuman performance, winning 100–0 against the previously published, champion-defeating AlphaGo. Starting from zero knowledge and without human data, AlphaGo Zero was able to teach itself to play Go and to develop novel strategies that provide new insights into the oldest of games. To beat world champions at the game of Go, the computer program AlphaGo has relied largely on supervised learning from millions of human expert moves. David Silver and colleagues have now produced a system called AlphaGo Zero, which is based purely on reinforcement learning and learns solely from self-play. Starting from random moves, it can reach superhuman level in just a couple of days of training and five million games of self-play, and can now beat all previous versions of AlphaGo. Because the machine independently discovers the same fundamental principles of the game that took humans millennia to conceptualize, the work suggests that such principles have some universal character, beyond human bias.

/pdf/mastering-the-game-of-go-without-human-knowledge-19y9mw638s.pdf

Mastering the game of Go without human knowledge

Diﬀusion-weighted magnetic resonance imaging is sensitive to the microstructural properties of brain tissue. However, estimating clinically and scientiﬁcally relevant microstructural properties from the measured signals remains a highly challenging inverse problem. This paper presents a novel framework for estimating microstructural parameters using recently developed orientationally invariant spherical convolutional neural networks and eﬃciently simulated training data with a known ground truth. The network was trained to predict the ground-truth parameter values from simulated noisy data and applied to imaging data acquired in a clinical setting to generate microstructural parameter maps. Our model could estimate model parameters from spherical data more accurately than conventional non-linear least squares or a multi-layer perceptron applied on powder-averaged data ( i.e. , the spherical mean technique, a popular method for orientationally invariant microstructural parameter estimation). Importantly, our method is generalizable and can be used to estimate the parameters of any Gaussian compartment model.

Microstructural neuroimaging using spherical convolutional neural networks

Diﬀusion-weighted magnetic resonance imaging is sensitive to the microstructural properties of brain tissue. However, estimating clinically and scientiﬁcally relevant microstructural properties from the measured signals remains a highly challenging inverse problem that deep learning may help solve. This study inves-tigated if recently developed orientationally invariant spherical convolutional neural networks can improve microstructural parameter estimation. A spherical convolutional neural network was trained to predict the ground-truth parameter values from simulated noisy data and applied to imaging data acquired in a clinical setting to generate microstructural parameter maps. The spherical convolutional neural network was more accurate and less orientationally variant than the benchmark methods (multi-layer perceptrons and the spherical mean technique). Our results show that spherical convolutional neural networks can be a compelling alternative to predicting parameters from powder-averaged data (i.e., data averaged over the acquired diﬀusion encoding directions). While we focused on constrained two- and three-compartment models of neuronal tissue, the presented network and training pipeline are generalizable and can be used to estimate the parameters of other Gaussian compartment models.

Microstructural parameter estimation using spherical convolutional neural networks

Communicating with humans is challenging for AIs because it requires a shared understanding of the world, complex semantics (e.g., metaphors or analogies), and at times multi-modal gestures (e.g., pointing with a finger, or an arrow in a diagram). We investigate these challenges in the context of Iconary, a collaborative game of drawing and guessing based on Pictionary, that poses a novel challenge for the research community. In Iconary, a Guesser tries to identify a phrase that a Drawer is drawing by composing icons, and the Drawer iteratively revises the drawing to help the Guesser in response. This back-and-forth often uses canonical scenes, visual metaphor, or icon compositions to express challenging words, making it an ideal test for mixing language and visual/symbolic communication in AI. We propose models to play Iconary and train them on over 55,000 games between human players. Our models are skillful players and are able to employ world knowledge in language models to play with words unseen during training.

Iconary: A Pictionary-Based Game for Testing Multimodal Communication with Drawings and Text

Brain injury is a common complication of sickle cell anaemia (SCA). White matter (WM) and cortical and subcortical grey matter (GM), structures may have reduced volume in patients with SCA. This study focuses on whether silent cerebral infarction (SCI), vasculopathy or anaemia affects WM and regional GM volumes in children living in Africa. Children with SCA (n = 144; aged 5–20 years; 74 male) and sibling controls (n = 53; aged 5–17 years; 29 male) underwent magnetic resonance imaging. Effects of SCI (n = 37), vasculopathy (n = 15), and haemoglobin were assessed. Compared with controls, after adjusting for age, sex and intracranial volume, patients with SCA had smaller volumes for WM and cortical, subcortical and total GM, as well as bilateral cerebellar cortex, globus pallidus, amygdala and right thalamus. Left globus pallidus volume was further reduced in patients with vasculopathy. Putamen and hippocampus volumes were larger in patients with SCA without SCI or vasculopathy than in controls. Significant positive effects of haemoglobin on regional GM volumes were confined to the controls. Patients with SCA generally have reduced GM volumes compared with controls, although some subcortical regions may be spared. SCI and vasculopathy may affect the trajectory of change in subcortical GM and WM volume. Brain volume in non‐SCA children may be vulnerable to contemporaneous anaemia.

/pdf/brain-volume-in-tanzanian-children-with-sickle-cell-anaemia-3j6h854i.pdf

Brain volume in Tanzanian children with sickle cell anaemia: A neuroimaging study

/pdf/correction-to-fibre-tract-segmentation-for-intraoperative-1roihvu1.pdf

Correction to: Fibre tract segmentation for intraoperative diffusion MRI in neurosurgical patients using tract-specific orientation atlas and tumour deformation modelling

Christopher Clark

Papers

Microstructural neuroimaging using spherical convolutional neural networks

Microstructural parameter estimation using spherical convolutional neural networks

Iconary: A Pictionary-Based Game for Testing Multimodal Communication with Drawings and Text

Brain volume in Tanzanian children with sickle cell anaemia: A neuroimaging study

Correction to: Fibre tract segmentation for intraoperative diffusion MRI in neurosurgical patients using tract-specific orientation atlas and tumour deformation modelling