Children acquire language subconsciously by observing the surrounding world and listening to descriptions. They can discover the meaning of words even without explicit language knowledge, and generalize to novel compositions effortlessly. In this paper, we bring this ability to AI, by studying the task of Visually grounded Language Acquisition (VLA). We propose a multimodal transformer model augmented with a novel mechanism for analogical reasoning, which approximates novel compositions by learning semantic mapping and reasoning operations from previously seen compositions. Our proposed method, Analogical Reasoning Transformer Networks (ARTNet), is trained on raw multimedia data (video frames and transcripts), and after observing a set of compositions such as "washing apple" or "cutting carrot", it can generalize and recognize new compositions in new video frames, such as "washing carrot" or "cutting apple". To this end, ARTNet refers to relevant instances in the training data and uses their visual features and captions to establish analogies with the query image. Then it chooses the suitable verb and noun to create a new composition that describes the new image best. Extensive experiments on an instructional video dataset demonstrate that the proposed method achieves significantly better generalization capability and recognition accuracy compared to state-of-the-art transformer models.

The perceived quality of a document is affected by various factors, including grammat- icality, readability, stylistics, and expertise depth, making the task of document quality assessment a complex one. In this paper, we explore this task in the context of assessing the quality of Wikipedia articles and academic papers. Observing that the visual rendering of a document can capture implicit quality indicators that are not present in the document text — such as images, font choices, and visual layout — we propose a joint model that combines the text content with a visual rendering of the document for document qual- ity assessment. Our joint model achieves state-of-the-art results over five datasets in two domains (Wikipedia and academic papers), which demonstrates the complementarity of textual and visual features, and the general applicability of our model. To examine what kinds of features our model has learned, we further train our model in a multi-task learning setting, where document quality assessment is the primary task and feature learning is an auxiliary task. Experimental results show that visual embeddings are better at learning structural features while textual embeddings are better at learning readability scores, which further verifies the complementarity of visual and textual features.

Feature evolvable learning has been widely studied in recent years where old features will vanish and new features will emerge when learning with streams. Conventional methods usually assume that a label will be revealed after prediction at each time step. However, in practice, this assumption may not hold whereas no label will be given at most time steps. To tackle this problem, we leverage the technique of manifold regularization to utilize the previous similar data to assist the refinement of the online model. Nevertheless, this approach needs to store all previous data which is impossible in learning with streams that arrive sequentially in large volume. Thus we need a buffer to store part of them. Considering that different devices may have different storage budgets, the learning approaches should be flexible subject to the storage budget limit. In this paper, we propose a new setting: Storage-Fit Feature-Evolvable streaming Learning (SF2EL) which incorporates the issue of rarely-provided labels into feature evolution. Our framework is able to fit its behavior to different storage budgets when learning with feature evolvable streams with unlabeled data. Besides, both theoretical and empirical results validate that our approach can preserve the merit of the original feature evolvable learning i.e., can always track the best baseline and thus perform well at any time step.

The objective of active learning (AL) is to train classification models with less number of labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.

Graph Neural Networks (GNNs) have recently caught great attention and achieved significant progress in graph-level applications. In order to handle graphs with different features and sizes, we propose a novel graph neural network, which we call HaarNet, to predict graph labels with interrelated convolution and pooling strategies. Similar to some existing routines, the model assembles unified graph-level representations from samples by first adopting graph convolutional layers to extract mutual information followed by graph pooling layers to downsample graph resolution. By a sequence of clusterings, we embed the intrinsic topological information of each graph into the GNN. Through the fast Haar transformation, we made our contribution to forming a smooth workflow that learns multi-scale graph representation with redundancy removed. As a result, our proposed framework obtains notable accuracy gains without sacrificing performance stability. Extensive experiments validate the superiority on graph classification and regression tasks, where our proposed HaarNet outperforms various existing GNN models, especially on big data sets.

The University of Florida on Tuesday announced a $70 million partnership that will bring artificial intelligence to the forefront of the school's technology programs and introduce the topic more broadly to the student body. The joint effort with the California-based company NVIDIA will result in the hiring of 100 new faculty and touch every UF graduate with at least one class exposing them to artificial intelligence concepts, the university said. It also will give UF the fastest artificial intelligence supercomputer in higher education, officials said. The discipline is a branch of computer science that has brought the world products like self-driving cars, food delivery robots and computers that engage humans in a game of chess. Artificial intelligence makes it possible for machines to perform human-like tasks by learning from experience and making adjustments.

TierPoint, a leading provider of secure, connected cloud and data center solutions at the edge of the internet, announced it will host the third annual BraveIT conference as a virtual event, the morning of September 16 and afternoon of September 17, with title sponsor support from Dell Technologies, Nutanix, and VMware. The opening keynote will be delivered by Tom Gruber, who led the design team responsible for Siri, the intelligent assistant. The closing keynote will be delivered by Dan Doctoroff, formerly the Deputy Mayor of New York City and now CEO of Sidewalk Labs, a leader in the development of smart cities. And, back by popular demand from last year's conference, Major General Brett Williams, USAF (Ret.) will help enterprise teams assess their Cyber IQ. Tom Gruber was cofounder, CTO, and head of design for the team that created Siri in 2011, bringing AI to the mainstream user experience.

Hosted by Dylan Doyle-Burke and Jessie J Smith, Radical AI is a podcast featuring the voices of the future in the field of artificial intelligence ethics. In this episode Jess and Dylan chat to Abeba Birhane about "Robot Rights? Should we grant robots rights? What is moral relationality and how can it be useful for designing machine learning algorithms? What is the algorithmic colonization of Africa and why is it harmful?

In this hands-on project, we will train a deep learning model to predict the type of food and then fine tune the model to improve its performance. This project could be practically applied in food industry to detect the type and quality of food. In this hands-on project, we will train a deep learning model to predict the type of food and then fine tune the model to improve its performance. This project could be practically applied in food industry to detect the type and quality of food.

Gabriel Guimaraes grew up in Vitória, Brazil, in a yellow house surrounded by star-fruit trees and chicken coops. His father, who wrote software for a local bank, instilled in him an interest in computers. On weekends, when Guimaraes got bored with Nintendo video games, he programmed his own. In grade school, he built a humanoid robot and wrote enough assembly code to make it zip around his home. In Vitória, an island city, his most ambitious peers dreamed of attending university in São Paulo, an hour away by plane.