Neural Networks (NNs) have been extensively used for a wide spectrum of real-world regression tasks, where the goal is to predict a numerical outcome such as revenue, effectiveness, or a quantitative result. In many such tasks, the point prediction is not enough, but also the uncertainty (i.e. risk, or confidence) of that prediction must be estimated. Standard NNs, which are most often used in such tasks, do not provide any such information. Existing approaches try to solve this issue by combining Bayesian models with NNs, but these models are hard to implement, more expensive to train, and usually do not perform as well as standard NNs. In this paper, a new framework called RIO is developed that makes it possible to estimate uncertainty in any pretrained standard NN. RIO models prediction residuals using Gaussian Process with a composite input/output kernel. The residual prediction and I/O kernel are theoretically motivated and the framework is evaluated in twelve real-world datasets. It is found to provide reliable estimates of the uncertainty, reduce the error of the point predictions, and scale well to large datasets. Given that RIO can be applied to any standard NN without modifications to model architecture or training pipeline, it provides an important ingredient in building real-world applications of NNs.

As deep learning applications continue to become more diverse, an interesting question arises: Can general problem solving arise from jointly learning several such diverse tasks? To approach this question, deep multi-task learning is extended in this paper to the setting where there is no obvious overlap between task architectures. The idea is that any set of (architecture,task) pairs can be decomposed into a set of potentially related subproblems, whose sharing is optimized by an efficient stochastic algorithm. The approach is first validated in a classic synthetic multi-task learning benchmark, and then applied to sharing across disparate architectures for vision, NLP, and genomics tasks. It discovers regularities across these domains, encodes them into sharable modules, and combines these modules systematically to improve performance in the individual tasks. The results confirm that sharing learned functionality across diverse domains and architectures is indeed beneficial, thus establishing a key ingredient for general problem solving in the future.

After 60 years, Artificial intelligence (AI) has moved from academic research discipline to a technology that affects people's lives every day. We have digital assistants with which you can carry rudimentary conversations, systems that make medical diagnoses more accurately than humans, and cars that drive themselves in regular traffic, for instance. At the same time, despite decades of development, AI is still in its infancy when it comes to commercial applications. There are few standards, little cooperation across companies and countries, and business users and consumers still rely on a small group of experts to be able contribute to AI solutions. There are significant issues that also need to be solved to ensure that as AI adoption grows, it creates positive effects on businesses and society.

As the complexity of neural network models has grown, it has become increasingly important to optimize their design automatically through metalearning. Methods for discovering hyperparameters, topologies, and learning rate schedules have lead to significant increases in performance. This paper shows that loss functions can be optimized with metalearning as well, and result in similar improvements. The method, Genetic Loss-function Optimization (GLO), discovers loss functions de novo, and optimizes them for a target task. Leveraging techniques from genetic programming, GLO builds loss functions hierarchically from a set of operators and leaf nodes. These functions are repeatedly recombined and mutated to find an optimal structure, and then a covariance-matrix adaptation evolutionary strategy (CMA-ES) is used to find optimal coefficients. Networks trained with GLO loss functions are found to outperform the standard cross-entropy loss on standard image classification tasks. Training with these new loss functions requires fewer steps, results in lower test error, and allows for smaller datasets to be used. Loss-function optimization thus provides a new dimension of metalearning, and constitutes an important step towards AutoML.

In the last decade or so we have seen tremendous progress in Artificial Intelligence (AI). AI is now in the real world, powering applications that have a large practical impact. Most of it is based on modeling, i.e. machine learning of statistical models that make it possible to predict what the right decision might be in future situations. The next step for AI is machine creativity, i.e. tasks where the correct, or even good, solutions are not known, but need to be discovered. Methods for machine creativity have existed for decades. I believe we are now in a similar situation as deep learning was a few years ago: with the million-fold increase in computational power, those methods can now be used to scale up to creativity in real-world tasks. In particular, Evolutionary Computation is in a unique position to take advantage of that power, and become the next deep learning.

Multitask learning, i.e. learning several tasks at once with the same neural network, can improve performance in each of the tasks. Designing deep neural network architectures for multitask learning is a challenge: There are many ways to tie the tasks together, and the design choices matter. The size and complexity of this problem exceeds human design ability, making it a compelling domain for evolutionary optimization. Using the existing state of the art soft ordering architecture as the starting point, methods for evolving the modules of this architecture and for evolving the overall topology or routing between modules are evaluated in this paper. A synergetic approach of evolving custom routings with evolved, shared modules for each task is found to be very powerful, significantly improving the state of the art in the Omniglot multitask, multialphabet character recognition domain. This result demonstrates how evolution can be instrumental in advancing deep neural network and complex system design in general.

As a classic example of imperfect information games, Heads-Up No-limit Texas Holdem (HUNL), has been studied extensively in recent years. While state-of-the-art approaches based on Nash equilibrium have been successful, they lack the ability to model and exploit opponents effectively. This paper presents an evolutionary approach to discover opponent models based Long Short Term Memory neural networks and on Pattern Recognition Trees. Experimental results showed that poker agents built in this method can adapt to opponents they have never seen in training and exploit weak strategies far more effectively than Slumbot 2017, one of the cutting-edge Nash-equilibrium-based poker agents. In addition, agents evolved through playing against relatively weak rule-based opponents tied statistically with Slumbot in heads-up matches. Thus, the proposed approach is a promising new direction for building high-performance adaptive agents in HUNL and other imperfect information games.

Deep multitask learning boosts performance by sharing learned structure across related tasks. This paper adapts ideas from deep multitask learning to the setting where only a single task is available. The method is formalized as pseudo-task augmentation, in which models are trained with multiple decoders for each task. Pseudo-tasks simulate the effect of training towards closely-related tasks drawn from the same universe. In a suite of experiments, pseudo-task augmentation is shown to improve performance on single-task learning problems. When combined with multitask learning, further improvements are achieved, including state-of-the-art performance on the CelebA dataset, showing that pseudo-task augmentation and multitask learning have complementary value. All in all, pseudo-task augmentation is a broadly applicable and efficient way to boost performance in deep learning systems.

Existing deep multitask learning (MTL) approaches align layers shared between tasks in a parallel ordering. Such an organization significantly constricts the types of shared structure that can be learned. The necessity of parallel ordering for deep MTL is first tested by comparing it with permuted ordering of shared layers. The results indicate that a flexible ordering can enable more effective sharing, thus motivating the development of a soft ordering approach, which learns how shared layers are applied in different ways for different tasks. Deep MTL with soft ordering outperforms parallel ordering methods across a series of domains. These results suggest that the power of deep MTL comes from learning highly general building blocks that can be assembled to meet the demands of each task.

Conversion rate optimization (CRO) means designing an e-commerce web interface so that as many users as possible take a desired action such as registering for an account, requesting a contact, or making a purchase. Such design is usually done by hand, evaluating one change at a time through A/B testing, or evaluating all combinations of two or three variables through multivariate testing. Traditional CRO is thus limited to a small fraction of the design space only. This paper describes Sentient Ascend, an automatic CRO system that uses evolutionary search to discover effective web interfaces given a human-designed search space. Design candidates are evaluated in parallel on line with real users, making it possible to discover and utilize interactions between the design elements that are difficult to identify otherwise. A commercial product since September 2016, Ascend has been applied to numerous web interfaces across industries and search space sizes, with up to four-fold improvements over human design. Ascend can therefore be seen as massively multivariate CRO made possible by AI.