Back in the days, computers simply carried out tasks from a set of instructions given to them. Now, with the immense advancements in artificial intelligence (AI), computers can now learn by example without human intervention with deep learning software. Hence, this is the reason behind the rise in popularity of deep learning applications. Deep learning is a promising and lucrative space that has achieved results that were thought to be impossible. It is providing many industries with innovative tools and valuable applications. For those who want to understand deep learning better, there are many available resources.

Back in the days, computers simply carried out tasks from a set of instructions given to them. Now, with the immense advancements in artificial intelligence (AI), computers can now learn by example without human intervention with deep learning software. Hence, this is the reason behind the rise in popularity of deep learning applications. Deep learning is a promising and lucrative space that has achieved results that were thought to be impossible. It is providing many industries with innovative tools and valuable applications. For those who want to understand deep learning better, there are many available resources.

In this post, we will learn the intuition behind the working of LSTM and GRU.

Chemical fingerprints [1] have long been the representation used to represent chemical structures as numbers, which are suitable inputs to machine learning models. A brief summary of chemical fingerprints is provided in another of my blog posts here. Above, we computed the fingerprint for Atorvastatin, a drug which generated over $100B in revenue over 2003–2013. At some point a few years ago, people started to realize [3] that instead of computing a non-differentiable fingerprint, we can compute a differentiable fingerprint. Then, by backpropagation, we can train not only a deep-learning model but also train the fingerprint-generating function itself. The promise would be to learn richer molecular representations.

Adaptive signal control system (ASCS) is the most advanced t raffic signal technology that regulates the signal phasing and timings considering the traffic patterns in real-time in order to reduce traffic congestion. Real-time prediction of traffic queue length can be used to adj ust the signal phasing and timings for different traffic movements at a signalized intersection with A SCS. The accuracy of the queue length prediction model varies based on the many factors, such as th e stochastic nature of the vehicle arrival rates at an intersection, time of the day, weather and driver characteristics. In addition, accurate queue length prediction for multilane, undersaturated and satur ated traffic scenarios at signalized intersections is challenging. Thus, the objective of this study is to devel op short-term queue length prediction models for signalized intersections that can be leveraged by adapt ive traffic signal control systems using four variations of Grey systems: (i) the first order single variab le Grey model (GM(1,1)); (ii) GM(1,1) with Fourier error corrections (EGM); (iii) the Grey Verhulst mo del (GVM), and (iv) GVM with Fourier error corrections (EGVM). The efficacy of the Grey models is th at they facilitate fast processing; as these models do not require a large amount of data; as would be needed in artificial intelligence models; and they are able to adapt to stochastic changes, unlike stat istical models. We have conducted a case study using queue length data from five intersections with ad aptive traffic signal control on a calibrated roadway network in Lexington, South Carolina. Grey models w ere compared with linear, nonlinear time series models, and long short-term memory (LSTM) neura l network. Based on our analyses, we found that EGVM reduces the prediction error over closest co mpeting models (i.e., LSTM and Additive Autoregressive (AAR) time series models) in predicting ave rage and maximum queue lengths by 40% and 42%, respectively, in terms of Root Mean Squared Error (R MSE), and 51% and 50%, respectively, in terms of Mean Absolute Error (MAE).

We present a continuous formulation of machine learning, as a problem in the calculus of variations and differential-integral equations, very much in the spirit of classical numerical analysis and statistical physics. We demonstrate that conventional machine learning models and algorithms, such as the random feature model, the shallow neural network model and the residual neural network model, can all be recovered as particular discretizations of different continuous formulations. We also present examples of new models, such as the flow-based random feature model, and new algorithms, such as the smoothed particle method and spectral method, that arise naturally from this continuous formulation. We discuss how the issues of generalization error and implicit regularization can be studied under this framework.

Recommender Systems (RS) have became a popular research topic and, since 2016, Deep Learning methods and techniques have been increasingly explored in this area. News RS are aimed to personalize users experiences and help them discover relevant articles from a large and dynamic search space. The main contribution of this research was named CHAMELEON, a Deep Learning meta-architecture designed to tackle the specific challenges of news recommendation. It consists of a modular reference architecture which can be instantiated using different neural building blocks. As information about users' past interactions is scarce in the news domain, the user context can be leveraged to deal with the user cold-start problem. Articles' content is also important to tackle the item cold-start problem. Additionally, the temporal decay of items (articles) relevance is very accelerated in the news domain. Furthermore, external breaking events may temporally attract global readership attention, a phenomenon generally known as concept drift in machine learning. All those characteristics are explicitly modeled on this research by a contextual hybrid session-based recommendation approach using Recurrent Neural Networks. The task addressed by this research is session-based news recommendation, i.e., next-click prediction using only information available in the current user session. A method is proposed for a realistic temporal offline evaluation of such task, replaying the stream of user clicks and fresh articles being continuously published in a news portal. Experiments performed with two large datasets have shown the effectiveness of the CHAMELEON for news recommendation on many quality factors such as accuracy, item coverage, novelty, and reduced item cold-start problem, when compared to other traditional and state-of-the-art session-based recommendation algorithms.

A BSTRACT Evolutionary-based optimization approaches have recently shown promising results in domains such as Atari and robot locomotion but less so in solving 3D tasks directly from pixels. This paper presents a method called Deep Innovation Protection (DIP) that allows training complex world models end-to-end for such 3D environments. The main idea behind the approach is to employ multiobjective optimization to temporally reduce the selection pressure on specific components in a world model, allowing other components to adapt. We investigate the emergent representations of these evolved networks, which learn a model of the world without the need for a specific forward-prediction loss. 1 I NTRODUCTION The ability of the brain to model the world arose from the process of evolution. It evolved because it helped organisms to survive and strive in their particular environments and not because such forward prediction was explicitly optimized for. In contrast to the emergent neural representations in nature, current world model approaches are often directly rewarded for their ability to predict future states of the environment (Schmidhuber, 1990; Ha & Schmidhuber, 2018; Hafner et al., 2018; Wayne et al., 2018). While it is undoubtedly useful to be able to explicitly encourage a model to predict what will happen next, in this paper we are interested in what type of representations can emerge from the less directed process of artificial evolution and what ingredients might be necessary to encourage the emergence of such predictive abilities. In particular, we are building on the recently introduced world model architecture introduced by Ha & Schmidhuber (2018). This agent model contains three different components: (1) a visual module, mapping high-dimensional inputs to a lower-dimensional representative code, (2) an LSTM-based memory component, and (3) a controller component that takes input from the visual and memory module to determine the agent's next action.

There is an analogy between machine learning systems and economic entities in that they are both adaptive, and their behaviour is specified in a more-or-less explicit way. It appears that the area of AI that is most analogous to the behaviour of economic entities is that of morally good decision-making, but it is an open question as to how precisely moral behaviour can be achieved in an AI system. This paper explores the analogy between these two complex systems, and we suggest that a clearer understanding of this apparent analogy may help us forward in both the socio-economic domain and the AI domain: known results in economics may help inform feasible solutions in AI safety, but also known results in AI may inform economic policy. If this claim is correct, then the recent successes of deep learning for AI suggest that more implicit specifications work better than explicit ones for solving such problems.

Learning predictive models from interaction with the world allows an agent, such as a robot, to learn about how the world works, and then use this learned model to plan coordinated sequences of actions to bring about desired outcomes. However, learning a model that captures the dynamics of complex skills represents a major challenge: if the agent needs a good model to perform these skills, it might never be able to collect the experience on its own that is required to learn these delicate and complex behaviors. Instead, we can imagine augmenting the training set with observational data of other agents, such as humans. Such data is likely more plentiful, but represents a different embodiment. F or example, videos of humans might show a robot how to use a tool, but (i) are not annotated with suitable robot actions, and (ii) contain a systematic distributional shift due to the embodiment differences between humans and robots. W e address the first challenge by formulating the corresponding graphical model and treating the action as an observed variable for the interaction data and an unobserved variable for the observation data, and the second challenge by using a domain-dependent prior . In addition to interaction data, our method is able to leverage videos of passive observations in a driving dataset and a dataset of robotic manipulation videos. A robotic planning agent equipped with our method can learn to use tools in a tabletop robotic manipulation setting by observing humans without ever seeing a robotic video of tool use.