This is a single day course from 9:00am to 2:00pm. You will need to bring your laptop and have python, TensoFflow 1.0 and pandas installed before the class. You can find the instructions here. If you have any difficulties let us know before the day of the training and we will provide you with support. We will be running two parallel sessions, one for new users who have minimal or no experience with TensorFlow and another one for advanced users.

When you think of sports analysis, you probably think of raw stats like time in the opposing half or shots on goal. However, that doesn't really tell teams how they should have played beyond vague suggestions. Researchers at Disney, Caltech and STATS believe they can do better: they've developed a system that uses deep learning to analyze athletes' decision-making processes. After enough training based on players' past actions, the system's neural networks can predict future moves and create a "ghost" of a player's typical performance. If a team flubbed a play, it could compare the real action against the predictive ghosts of more effective teams to see how players should have acted.

We propose Edward, a Turing-complete probabilistic programming language. Edward defines two compositional representations---random variables and inference. By treating inference as a first class citizen, on a par with modeling, we show that probabilistic programming can be as flexible and computationally efficient as traditional deep learning. For flexibility, Edward makes it easy to fit the same model using a variety of composable inference methods, ranging from point estimation to variational inference to MCMC. In addition, Edward can reuse the modeling representation as part of inference, facilitating the design of rich variational models and generative adversarial networks. For efficiency, Edward is integrated into TensorFlow, providing significant speedups over existing probabilistic systems. For example, we show on a benchmark logistic regression task that Edward is at least 35x faster than Stan and 6x faster than PyMC3. Further, Edward incurs no runtime overhead: it is as fast as handwritten TensorFlow.

We propose the spatiotemporal estimation of the demand that is a function of variable effects related to traffic, pricing and weather conditions. With respect to the methodology, a single decision tree, bootstrap-aggregated (bagged) decision trees, random forest, boosted decision trees, and artificial neural network for regression have been adapted and systematically compared using various statistics, e.g. R-square, Root Mean Square Error (RMSE), and slope. To better assess the quality of the models, they have been tested on a real case study using the data of DiDi Chuxing, the main on-demand ride-hailing service provider in China. In the current study, 199,584 time-slots describing the spatiotemporal ride-hailing demand has been extracted with an aggregated-time interval of 10 mins. All the methods are trained and validated on the basis of two independent samples from this dataset. The results revealed that boosted decision trees provide the best prediction accuracy (RMSE 16.41), while avoiding the risk of over-fitting, followed by artificial neural network (20.09), random forest (23.50), bagged decision trees (24.29) and single decision tree (33.55).

Bird sounds possess distinctive spectral structure which may exhibit small shifts in spectrum depending on the bird species and environmental conditions. In this paper, we propose using convolutional recurrent neural networks on the task of automated bird audio detection in real-life environments. In the proposed method, convolutional layers extract high dimensional, local frequency shift invariant features, while recurrent layers capture longer term dependencies between the features extracted from short time frames. This method achieves 88.5% Area Under ROC Curve (AUC) score on the unseen evaluation data and obtains the second place in the Bird Audio Detection challenge.

Bayesian optimization has become a successful tool for hyperparameter optimization of machine learning algorithms, such as support vector machines or deep neural networks. Despite its success, for large datasets, training and validating a single configuration often takes hours, days, or even weeks, which limits the achievable performance. To accelerate hyperparameter optimization, we propose a generative model for the validation error as a function of training set size, which is learned during the optimization process and allows exploration of preliminary configurations on small subsets, by extrapolating to the full dataset. We construct a Bayesian optimization procedure, dubbed Fabolas, which models loss and training time as a function of dataset size and automatically trades off high information gain about the global optimum against computational cost. Experiments optimizing support vector machines and deep neural networks show that Fabolas often finds high-quality solutions 10 to 100 times faster than other state-of-the-art Bayesian optimization methods or the recently proposed bandit strategy Hyperband.

The Toronto Raptors already have a manual ghosting system where coaches mark out where they think players should have been. It can create ghosts in real time, even in soccer (aka football) and other sports where the continuous play can lead to predictions that gradually veer from realistic outcomes. The scientists rely on imitation learning, where AI bases its actions on demonstrations, to keep that long-term prediction in check. The early results are promising. In an example soccer match between Fulham and Swansea, a league-average ghost team replacing Swansea performed about as well in a defensive situation... not well at all, unfortunately.

I recently stumbled across a research paper, Using Deep Learning and Google Street View to Estimate the Demograp..., which piqued my interest in derivative uses of data, an ongoing research interest of mine. A variety of deep learning techniques were used to draw conclusions about relationships of car ownership, political affiliation and demographics. For those headline skimmers, you may be led to believe that researchers have just uncovered a vastly cheaper and more timely approach to perform the national census and make predictive claims about the population. The researchers' contention that official statistics are expensive and lagging is spot on. The principal US unemployment survey is performed in person or via telephone.

Sports analytics is routinely used to assign values to such things as shots taken or to compare player performance, but a new automated method based on deep learning techniques - developed by researchers at Disney Research, California Institute of Technology and STATS, a supplier of sports data - will provide coaches and teams with a quicker tool to help assess defensive athletic performance in any game situation. The innovative method analyzes detailed game data on player and ball positions to create models, or "ghosts," of how a typical player in a league or on another team would behave when an opponent is on the attack. It is then possible to visually compare what a team's players actually did during a defensive play versus what the ghost players would have done. "With the innovation of data-driven ghosting, we can now, for the first time, scalably quantify, analyze and compare detailed defensive behavior," said Peter Carr, research scientist at Disney Research. "Despite what skeptics might say, you can indeed measure defense."

This is a curated list of the most cited deep learning papers (since 2012) posted by Terry Taewoong Um. Source for picture: What is deep learning and how does it work?