Reinforcement Learning (RL) is a learning paradigm concerned with learning to control a system so as to maximize an objective over the long term. This approach to learning has received immense interest in recent times and success manifests itself in the form of human-level performance on games like \textit{Go}. While RL is emerging as a practical component in real-life systems, most successes have been in Single Agent domains. This report will instead specifically focus on challenges that are unique to Multi-Agent Systems interacting in mixed cooperative and competitive environments. The report concludes with advances in the paradigm of training Multi-Agent Systems called \textit{Decentralized Actor, Centralized Critic}, based on an extension of MDPs called \textit{Decentralized Partially Observable MDP}s, which has seen a renewed interest lately.

Stochastic optimization is one of the effective approach to deal with the large-scale machine learning problems and the recent research has focused on reduction of variance, caused by the noisy approximations of the gradients, and momentum acceleration. In this paper, we have proposed simple variants of SAAG-I and II (Stochastic Average Adjusted Gradient) \cite{Chauhan2017Saag}, called SAAG-III and IV, respectively. Unlike SAAG-I, starting point is set to average of previous epoch in SAAG-III, and unlike SAAG-II, the snap point and starting point are set to average and last iterate of previous epoch, respectively. To determine the step size, we introduce Stochastic Backtracking-Armijo line Search (SBAS) which performs line search only on selected mini-batch of data points. Since backtracking line search is not suitable for large-scale problems and the constants used to find the step size, like Lipschitz constant, are not always available so SBAS could be very effective in such cases. We also extend SAAGs (I, II, III, IV), to solve non-smooth problems and design two update rules for smooth and non-smooth problems. Moreover, our theoretical results prove linear convergence of SAAG-IV for all the four combinations of smoothness and strong-convexity, in expectation. Finally, our experimental studies prove the efficacy of proposed methods against the state-of-art techniques, like, SVRG and VR-SGD.

In this project, we developed a deep learning system applied to human retina images for medical diagnostic decision support. The retina images were provided by EyePACS (Eyepacs, LLC). These images were used in the framework of a Kaggle contest (Kaggle INC, 2017), whose purpose to identify diabetic retinopathy signs through an automatic detection system. Using as inspiration one of the solutions proposed in the contest, we implemented a model that successfully detects diabetic retinopathy from retina images. After a carefully designed preprocessing, the images were used as input to a deep convolutional neural network (CNN). The CNN performed a feature extraction process followed by a classification stage, which allowed the system to differentiate between healthy and ill patients using five categories. Our model was able to identify diabetic retinopathy in the patients with an agreement rate of 76.73% with respect to the medical expert's labels for the test data.

Cooperation of a large number of possibly small-sized robots, called robotic swarm, will play a significant role in complex missions that existing operational concepts using a few large robots could not deal with [1]. Even if every single robot (or called agent) in a swarm is incapable of accomplishing a task alone, their cooperation will lead to successful outcomes [2]-[5]. The possible applications include environmental monitoring [6], ad-hoc network relay [7], disaster management [8], cooperative radar jamming [9], to name a few. Due to the large cardinality of a swarm robot system, however, it is infeasible for human operators to supervise each agent directly, but needed to entrust the swarm with certain levels of decision-makings (e.g., task allocation, path planning, and individual control). Thereby, what only remains is to provide a high-level mission description, which is manageable for a few or even a single human operator. Nevertheless, there still exist various challenges in the autonomous decisionmaking of robotic swarms. Among them, this paper addresses a task allocation problem where the number of agents is higher than that of tasks: how to partition a set of agents into subgroups and assign the subgroups to each task.

Researchers hoping to deploy a 600-meter plastic-sweeper to the Pacific Ocean to clean up the notorious floating Great Garbage Patch have revealed the final design for their contraption. The gigantic'pac man' system consists of a 600-meter-long floating tube that sits at the surface of the water, with a tapered 3-meter-deep skirt attached below to catch plastic waste. It harnesses the power of wind and surface waves to autonomously sweep through the area, gathering up plastic waste as it goes. The gigantic'pac man' system consists of a 600-meter-long floater that sits at the surface of the water, with a tapered 3-meter-deep skirt attached below to catch plastic waste. In the water, the'pac man' will catch plastic in a skirt, which will be emptired by a boat every few weeks Ocean Cleanup Project was forced to radically redesign the system after tests of their original system found it moved too much due to waves.

A former Uber and Lyft driver has been dropped from both platforms after it was revealed that he secretly livestreamed his passengers online. Jason Gargac, a 32-year-old aspiring police officer from Missouri, would regularly broadcast his rides to thousands of subscribers on the livestreaming service Twitch, according to local reports. His riders – of which there were reportedly more than 700 in the last four months – rarely knew they were being broadcast. I'm embarrassed," one passenger told the St Lous Post-Dispatch, who first reported the story. "We got in an Uber at 2am to be safe, and then I find out that because of that, everything I said in that car is online and people are watching me.

Russia is planning to blast two robot astronauts into space to work on the international space station. Scientists have developed the advanced machines, named FEDOR, to conduct rescues - even though they have recently been recently trained to use firearms. According to RIA Novosti, the robots could be blasted into space as soon as August 2019. Unlike previous robots, DefenseOne.com reports that these will be sent into orbit as crew members on board the Soyuz rocket and not placed into the hold. However, no humans will be on board during the launch.

A demonstration video that veteran University College, London neuroscientist John O'Keefe often presents in lectures shows a rat moving around the inside of a box. Every time the rat heads for the top-left corner, loud pops play through a speaker; those sounds are the result of the firing of a specific neuron attached to an electrode. The neuron only fires when the rat moves to the same small area of the box. This connection of certain neurons to locations led O'Keefe and student Jonathon Dostrovsky to name those neurons "place cells" when they encountered the phenomenon in the early 1970s. Today, researchers such as Huajin Tang, director of the Neuromorphic Computing Research Center at Sichuan University, China, are using maps of computer memory to demonstrate how simulated neurons fire in much the same way inside one of their wheeled robots.

There are lots of reasons why colonizing space seems compelling. The popular astronomer Neil deGrasse Tyson argues that it would stimulate the economy and inspire the next generation of scientists. Elon Musk, who founded SpaceX, argues that "there is a strong humanitarian argument for making life multiplanetary…to safeguard the existence of humanity in the event that something catastrophic were to happen." And the late astrophysicist Stephen Hawking has conjectured that if humanity fails to colonize space within 100 years, we could face extinction. To be sure, humanity will eventually need to escape Earth to survive, since the sun will make the planet uninhabitable in about 1 billion years.

We think the future workplace will be one that's led by people who excel at finding and acting quickly with their information. But today, our work is fractured. Millions of unnecessary steps stand between us and the information we need to do our jobs. Think of Journal as a companion app that connects your tools. It helps you get to all your information fast so you can use your time and energy on the things that matter. Join our waitlist, and we'll give you early access. "How should brands use future technologies?"