Machine Learning


Biasing MCTS with Features for General Games

arXiv.org Artificial Intelligence

This paper proposes using a linear function approximator, rather than a deep neural network (DNN), to bias a Monte Carlo tree search (MCTS) player for general games. This is unlikely to match the potential raw playing strength of DNNs, but has advantages in terms of generality, interpretability and resources (time and hardware) required for training. Features describing local patterns are used as inputs. The features are formulated in such a way that they are easily interpretable and applicable to a wide range of general games, and might encode simple local strategies. We gradually create new features during the same self-play training process used to learn feature weights. We evaluate the playing strength of an MCTS player biased by learnt features against a standard upper confidence bounds for trees (UCT) player in multiple different board games, and demonstrate significantly improved playing strength in the majority of them after a small number of self-play training games.


Short-Term Prediction and Multi-Camera Fusion on Semantic Grids

arXiv.org Artificial Intelligence

An environment representation (ER) is a substantial part of every autonomous system. It introduces a common interface between perception and other system components, such as decision making, and allows downstream algorithms to deal with abstracted data without knowledge of the used sensor. In this work, we propose and evaluate a novel architecture that generates an egocentric, grid-based, predictive, and semantically-interpretable ER. In particular, we provide a proof of concept for the spatio-temporal fusion of multiple camera sequences and short-term prediction in such an ER. Our design utilizes a strong semantic segmentation network together with depth and egomotion estimates to first extract semantic information from multiple camera streams and then transform these separately into egocentric temporally-aligned bird's-eye view grids. A deep encoder-decoder network is trained to fuse a stack of these grids into a unified semantic grid representation and to predict the dynamics of its surrounding. We evaluate this representation on real-world sequences of the Cityscapes dataset and show that our architecture can make accurate predictions in complex sensor fusion scenarios and significantly outperforms a model-driven baseline in a category-based evaluation.


Nvidia unveils incredible 'smart paintbrush' software that uses AI to turn simple doodles into art

Daily Mail

A new piece of software developed by American tech company, NVIDIA, uses deep-learning to elevate even the roughest sketches into works of art. The new program, dubbed GauGAN, after famous French impressionist Paul Gaugin, uses a tool called generative adversarial networks (GAN) to interpret simple lines and convert them into hyper-realistic images. Its application could help professionals across a range of disciplines such as architecture and urban planning render images and visualizations faster and with greater accuracy, according to the company. A new piece of software developed by American tech company, NVIDIA, uses deep-learning to elevate even the roughest sketches into works of art. Simple shapes become mountains and lakes with just a stroke of what NVIDIA calls a'smart paintbrush' Artificial intelligence systems rely on neural networks, which try to simulate the way the brain works in order to learn.


Women in Data Science conference unites global community of researchers and practitioners

MIT News

The MIT Institute for Data, Systems, and Society (IDSS) convened professional data scientists, academic researchers, and students from a variety of disciplines for the third annual daylong Women in Data Science (WiDS) conference in Cambridge. WiDS Cambridge is one of many global satellite events of the WiDS conference at Stanford University, where attendees join a global community of data science researchers and practitioners. The conference is open to anyone interested in data science, but strives especially to create opportunities for women in the field to showcase their work and network with each other. "I think WiDS is a great opportunity to bring together women at all professional levels -- students, postdocs, faculty, and professionals in industry -- who are working in data science, building community, and learning from a wide variety of perspectives," said Stefanie Jegelka, an IDSS affiliate faculty member with the Department of Electrical Engineering and Computer Science (EECS). Jegelka is an MIT WiDS planning committee member who also gave a talk exploring the properties of neural networks, focusing on ResNet architecture and neural networks for graphs.


Machine learning identifies links between world's oceans

MIT News

Oceanographers studying the physics of the global ocean have long found themselves facing a conundrum: Fluid dynamical balances can vary greatly from point to point, rendering it difficult to make global generalizations. Factors like the wind, local topography, and meteorological exchanges make it difficult to compare one area to another. To add to the complexity, one would have to analyze billions of data points for numerous parameters -- temperature, salinity, velocity, how things change with depth, whether there is a trend present -- to pinpoint what physics are most dominant in a given region. "You would have to look at an overwhelming number of different global maps and mentally match them up to figure out what matters most where," says Maike Sonnewald, a postdoc working in the MIT Department of Earth, Atmospheric and Planetary Sciences (EAPS) and a member of the EAPS Program in Atmospheres, Oceans and Climate (PAOC). Sonnewald, who has a background in physical oceanography and data science, uses computers to reveal connections and patterns in the ocean that would otherwise be beyond human capability.


This AI predicts online trolling before it happens

Mashable

How do you keep online trolls in check? Dr. Srijan Kumar, a post-doctoral research fellow in computer science at Stanford University, is developing an AI that predicts online conflict. His research uses data science and machine learning to promote healthy online interactions and curb deception, misbehavior, and disinformation. His work is currently deployed inside Indian e-commerce platform Flipkart, which uses it to spot fake reviewers. We spoke to Dr. Kumar ahead of a lecture on healthy online interactions at USC.


Artificial Intelligence : from Research to Application ; the Upper-Rhine Artificial Intelligence Symposium (UR-AI 2019)

arXiv.org Artificial Intelligence

The TriRhenaTech alliance universities and their partners presented their competences in the field of artificial intelligence and their cross-border cooperations with the industry at the tri-national conference 'Artificial Intelligence : from Research to Application' on March 13th, 2019 in Offenburg. The TriRhenaTech alliance is a network of universities in the Upper Rhine Trinational Metropolitan Region comprising of the German universities of applied sciences in Furtwangen, Kaiserslautern, Karlsruhe, and Offenburg, the Baden-Wuerttemberg Cooperative State University Loerrach, the French university network Alsace Tech (comprised of 14 'grandes \'ecoles' in the fields of engineering, architecture and management) and the University of Applied Sciences and Arts Northwestern Switzerland. The alliance's common goal is to reinforce the transfer of knowledge, research, and technology, as well as the cross-border mobility of students.


Empirical Evaluations of Seed Set Selection Strategies for Predictive Coding

arXiv.org Artificial Intelligence

Training documents have a significant impact on the performance of predictive models in the legal domain. Yet, there is limited research that explores the effectiveness of the training document selection strategy - in particular, the strategy used to select the seed set, or the set of documents an attorney reviews first to establish an initial model. Since there is limited research on this important component of predictive coding, the authors of this paper set out to identify strategies that consistently perform well. Our research demonstrated that the seed set selection strategy can have a significant impact on the precision of a predictive model. Enabling attorneys with the results of this study will allow them to initiate the most effective predictive modeling process to comb through the terabytes of data typically present in modern litigation. This study used documents from four actual legal cases to evaluate eight different seed set selection strategies. Attorneys can use the results contained within this paper to enhance their approach to predictive coding.


Engaging Image Captioning Via Personality

arXiv.org Artificial Intelligence

Standard image captioning tasks such as COCO and Flickr30k are factual, neutral in tone and (to a human) state the obvious (e.g., "a man playing a guitar"). While such tasks are useful to verify that a machine understands the content of an image, they are not engaging to humans as captions. With this in mind we define a new task, Personality-Captions, where the goal is to be as engaging to humans as possible by incorporating controllable style and personality traits. We collect and release a large dataset of 201,858 of such captions conditioned over 215 possible traits. We build models that combine existing work from (i) sentence representations (Mazare et al., 2018) with Transformers trained on 1.7 billion dialogue examples; and (ii) image representations (Mahajan et al., 2018) with ResNets trained on 3.5 billion social media images. We obtain state-of-the-art performance on Flickr30k and COCO, and strong performance on our new task. Finally, online evaluations validate that our task and models are engaging to humans, with our best model close to human performance.


Augmenting expert detection of early coronary artery occlusion from 12 lead electrocardiograms using deep learning

arXiv.org Artificial Intelligence

Early diagnosis of acute coronary artery occlusion based on electrocardiogram (ECG) findings is essential for prompt delivery of primary percutaneous coronary intervention. Current ST elevation (STE) criteria are specific but insensitive. Consequently, it is likely that many patients are missing out on potentially life-saving treatment. Experts combining non-specific ECG changes with STE detect ischaemia with higher sensitivity, but at the cost of specificity. We show that a deep learning model can detect ischaemia caused by acute coronary artery occlusion with a better balance of sensitivity and specificity than STE criteria, existing computerised analysers or expert cardiologists.