An important development in deep learning from the earliest MLPs has been a move towards architectures with structural inductive biases which enable the model to keep distinct sources of information and routes of processing well-separated. This structure is linked to the notion of independent mechanisms from the causality literature, in which a mechanism is able to retain the same processing as irrelevant aspects of the world are changed. For example, convnets enable separation over positions, while attention-based architectures (especially Transformers) learn which combination of positions to process dynamically. In this work we explore a way in which the Transformer architecture is deficient: it represents each position with a large monolithic hidden representation and a single set of parameters which are applied over the entire hidden representation. This potentially throws unrelated sources of information together, and limits the Transformer's ability to capture independent mechanisms. To address this, we propose Transformers with Independent Mechanisms (TIM), a new Transformer layer which divides the hidden representation and parameters into multiple mechanisms, which only exchange information through attention. Additionally, we propose a competition mechanism which encourages these mechanisms to specialize over time steps, and thus be more independent. We study TIM on a large-scale BERT model, on the Image Transformer, and on speech enhancement and find evidence for semantically meaningful specialization as well as improved performance.

Salient object detection (SOD) is viewed as a pixel-wise saliency modeling task by traditional deep learning-based methods. Although great progress has been made, a challenge of modern SOD models is the insufficient utilization of inter-pixel information, which usually results in imperfect segmentations near the edge regions. As we demonstrate, using a saliency map as the network output is a sub-optimal choice. To address this problem, we propose a connectivity-based approach named bilateral connectivity network (BiconNet), which uses a connectivity map instead of a saliency map as the network output for effective modeling of inter-pixel relationships and object saliency. Moreover, we propose a bilateral voting module to enhance the output connectivity map and a novel edge feature enhancement method that efficiently utilizes edge-specific features with negligible parameter increase. We show that our model can use any existing saliency-based SOD framework as its backbone. Through comprehensive experiments on five benchmark datasets, we demonstrate that our proposed method outperforms state-of-the-art SOD approaches.

Recent years have seen a proliferation of new digital products for the efficient management of indoor spaces, with important applications like emergency management, virtual property showcasing and interior design. These products rely on accurate 3D models of the environments considered, including information on both architectural and non-permanent elements. These models must be created from measured data such as RGB-D images or 3D point clouds, whose capture and consolidation involves lengthy data workflows. This strongly limits the rate at which 3D models can be produced, preventing the adoption of many digital services for indoor space management. We provide an alternative to such data-intensive procedures by presenting Walk2Map, a data-driven approach to generate floor plans only from trajectories of a person walking inside the rooms. Thanks to recent advances in data-driven inertial odometry, such minimalistic input data can be acquired from the IMU readings of consumer-level smartphones, which allows for an effortless and scalable mapping of real-world indoor spaces. Our work is based on learning the latent relation between an indoor walk trajectory and the information represented in a floor plan: interior space footprint, portals, and furniture. We distinguish between recovering area-related (interior footprint, furniture) and wall-related (doors) information and use two different neural architectures for the two tasks: an image-based Encoder-Decoder and a Graph Convolutional Network, respectively. We train our networks using scanned 3D indoor models and apply them in a cascaded fashion on an indoor walk trajectory at inference time. We perform a qualitative and quantitative evaluation using both simulated and measured, real-world trajectories, and compare against a baseline method for image-to-image translation. The experiments confirm the feasibility of our approach.

Graph Neural Networks (GNNs), a generalization of deep neural networks on graph data have been widely used in various domains, ranging from drug discovery to recommender systems. However, GNNs on such applications are limited when there are few available samples. Meta-learning has been an important framework to address the lack of samples in machine learning, and in recent years, the researchers have started to apply meta-learning to GNNs. In this work, we provide a comprehensive survey of different meta-learning approaches involving GNNs on various graph problems showing the power of using these two approaches together. We categorize the literature based on proposed architectures, shared representations, and applications. Finally, we discuss several exciting future research directions and open problems.

After the explosive growth of open source machine learning and deep learning frameworks, the field is more accessible than ever. Thanks to this, it went from a tool for researchers to a widely adopted and used method, fueling the insane growth of technology we experience now. Understanding how the algorithms really work can give you a huge advantage in designing, developing and debugging machine learning systems. Due to its mathematical nature, this task can seem daunting for many. However, this does not have to be the way.

What does Timnit Gebru's firing and the recent papers coming out of Google tell us about the state of research at the world's biggest AI research department. The high point for Google's research in to Artifical Intelligence may well turn out to be the 19th of October 2017. This was the date that David Silver and his co-workers at DeepMind published a report, in the journal Nature, showing how their deep-learning algorithm AlphaGo Zero was a better Go player than not only the best human in the world, but all other Go-playing computers. What was most remarkable about AlphaGo Zero was that it worked without human assistance. The researchers set up a neural network, let it play lots of games of Go against itself and a few days later it was the best Go player in the world. Then they showed it chess and it took only four hours to become the best chess player in the world.

The recent outbreak of COVID-19 has motivated researchers to contribute in the area of medical imaging using artificial intelligence and deep learning. Super-resolution (SR), in the past few years, has produced remarkable results using deep learning methods. The ability of deep learning methods to learn the non-linear mapping from low-resolution (LR) images to their corresponding high-resolution (HR) images leads to compelling results for SR in diverse areas of research. In this paper, we propose a deep learning based image super-resolution architecture in Tchebichef transform domain. This is achieved by integrating a transform layer into the proposed architecture through a customized Tchebichef convolutional layer (TCL).

Graph Neural Networks (GNNs) is a subtype of neural networks that operate on data structured as graphs. In an article covered earlier on Geometric Deep Learning, we saw how image processing, image classification, and speech recognition are represented in the Euclidean space. Graphs are non-Euclidean and can be used to study and analyse 3D data. GNN involves converting non-structured data like images and text into graphs to perform analysis. A graph is usually a representation of a data structure with two components -- Vertices (V) and Edges (E), which is generally put as; G Ω (V, E).

GPT-3 has created a lot of buzz since its release a few months ago. The system can generate (almost) plausible conversations with the likes of Nietzsche, write op eds for The Guardian and was even used successfully to post undercover comments on Reddit for a week. But even with GPT-3, AI is still stuck in Uncanny Valley. GPT-3 output feels like it was written by a human at first glance, but it isn't quite. On closer inspection, it lacks substance and coherence.

It's hard to feel connected to someone who's gone through a static photo. So a company called MyHeritage who provides automatic AI-powered photo enhancements is now offering a new service that can animate people in old photos creating a short video that looks like it was recorded while they posed and prepped for the portrait. Called Deep Nostalgia, the resulting videos are reminiscent of the Live Photos feature in iOS and iPadOS where several seconds of video are recorded and saved before and after the camera app's shutter is pressed. But where Live Photos is intended to be used to find the perfect shot and framing that may have been missed the exact second the shutter was pressed, Deep Nostalgia is instead meant to bring still shots, even those not captured on a modern smartphone, to life. The conversion process is completely automated. Users simply need to upload a photograph through the MyHeritage website where it's first sharpened and enhanced to not only improve the quality of the final animation but to also make it easier for the deep learning algorithm (created by a company called D-ID) to do its thing.