Much attention has been devoted recently to the generalization puzzle in deep learning: large, deep networks can generalize well, but existing theories bounding generalization error are exceedingly loose, and thus cannot explain this striking performance. Furthermore, a major hope is that knowledge may transfer across tasks, so that multi-task learning can improve generalization on individual tasks. However we lack analytic theories that can quantitatively predict how the degree of knowledge transfer depends on the relationship between the tasks. We develop an analytic theory of the nonlinear dynamics of generalization in deep linear networks, both within and across tasks. In particular, our theory provides analytic solutions to the training and testing error of deep networks as a function of training time, number of examples, network size and initialization, and the task structure and SNR. Our theory reveals that deep networks progressively learn the most important task structure first, so that generalization error at the early stopping time primarily depends on task structure and is independent of network size. This suggests any tight bound on generalization error must take into account task structure, and explains observations about real data being learned faster than random data. Intriguingly our theory also reveals the existence of a learning algorithm that proveably out-performs neural network training through gradient descent. Finally, for transfer learning, our theory reveals that knowledge transfer depends sensitively, but computably, on the SNRs and input feature alignments of pairs of tasks.

Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning problematic. Recently, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more meaningful comparisons, we identified three distinct continual learning scenarios based on whether task identity is known and, if it is not, whether it needs to be inferred. Performing the split and permuted MNIST task protocols according to each of these scenarios, we found that regularization-based approaches (e.g., elastic weight consolidation) failed when task identity needed to be inferred. In contrast, generative replay combined with distillation (i.e., using class probabilities as "soft targets") achieved superior performance in all three scenarios. In addition, we reduced the computational cost of generative replay by integrating the generative model into the main model by equipping it with generative feedback connections. This Replay-through-Feedback approach substantially shortened training time with no or negligible loss in performance. We believe this to be an important first step towards making the powerful technique of generative replay scalable to real-world continual learning applications.

Multiple query criteria active learning (MQCAL) methods have a higher potential performance than conventional active learning methods in which only one criterion is deployed for sample selection. A central issue related to MQCAL methods concerns the development of an integration criteria strategy (ICS) that makes full use of all criteria. The conventional ICS adopted in relevant research all facilitate the desired effects, but several limitations still must be addressed. For instance, some of the strategies are not sufficiently scalable during the design process, and the number and type of criteria involved are dictated. Thus, it is challenging for the user to integrate other criteria into the original process unless modifications are made to the algorithm. Other strategies are too dependent on empirical parameters, which can only be acquired by experience or cross-validation and thus lack generality; additionally, these strategies are counter to the intention of active learning, as samples need to be labeled in the validation set before the active learning process can begin. To address these limitations, we propose a novel MQCAL method for classification tasks that employs a third strategy via weighted rank aggregation. The proposed method serves as a heuristic means to select high-value samples of high scalability and generality and is implemented through a three-step process: (1) the transformation of the sample selection to sample ranking and scoring, (2) the computation of the self-adaptive weights of each criterion, and (3) the weighted aggregation of each sample rank list. Ultimately, the sample at the top of the aggregated ranking list is the most comprehensively valuable and must be labeled. Several experiments generating 257 wins, 194 ties and 49 losses against other state-of-the-art MQCALs are conducted to verify that the proposed method can achieve superior results.

We present a novel solution to the problem of simulation-to-real transfer, which builds on recent advances in robot skill decomposition. Rather than focusing on minimizing the simulation-reality gap, we learn a set of diverse policies that are parameterized in a way that makes them easily reusable. This diversity and parameterization of low-level skills allows us to find a transferable policy that is able to use combinations and variations of different skills to solve more complex, high-level tasks. In particular, we first use simulation to jointly learn a policy for a set of low-level skills, and a "skill embedding" parameterization which can be used to compose them. Later, we learn high-level policies which actuate the low-level policies via this skill embedding parameterization. The high-level policies encode how and when to reuse the low-level skills together to achieve specific high-level tasks. Importantly, our method learns to control a real robot in joint-space to achieve these high-level tasks with little or no on-robot time, despite the fact that the low-level policies may not be perfectly transferable from simulation to real, and that the low-level skills were not trained on any examples of high-level tasks. We illustrate the principles of our method using informative simulation experiments. We then verify its usefulness for real robotics problems by learning, transferring, and composing free-space and contact motion skills on a Sawyer robot using only joint-space control. We experiment with several techniques for composing pre-learned skills, and find that our method allows us to use both learning-based approaches and efficient search-based planning to achieve high-level tasks using only pre-learned skills.

Machine Learning (ML) is not easy, but creating a good workflow which you can reproduce, revisit and deploy to production is even harder. There has been many advances towards creating a good platform or managing solution for ML. Note that this is not the Data Science (DS) Lifecycle, which is more complex and has many parts. The ML lifecycle exists inside the DS lifecycle. These packages are great, but not so easy to follow.

The tech giant is celebrating its 20th birthday with a new Google Doodle marking the special occasion. The multi-billion dollar technology titan has certainly come a long way from its humble beginnings in the dorm rooms of Stanford University just two decades ago. Here's a look at the history of Google and what a Google Doodle is. Google's newest Doodle marks the company's milestone 20th birthday in September 2018 Google's history dates back to 1995 at Stanford University in Stanford, California after prospective graduate school student Larry Page met Sergey Brin, a student at the college assigned to show him around. After becoming friends, both Page and Brin developed a search engine from their dorm rooms known as Backrub in 1996 designed to improve online search by using links to determine the importance of website pages.

The Amazon SageMaker machine learning service is a full platform that greatly simplifies the process of training and deploying your models at scale. However, there are still major gaps to enabling data scientists to do research and development without having to go through the heavy lifting of provisioning the infrastructure and developing their own continuous delivery practices to obtain quick feedback. In this talk, you will learn how to leverage AWS CodePipeline, CloudFormation, CodeBuild, and SageMaker to create continuous delivery pipelines that allow the data scientist to use a repeatable process to build, train, test and deploy their models. Below, I've included a screencast of the talk I gave at the AWS NYC Summit in July 2018 along with a transcript (generated by Amazon Transcribe – another Machine Learning service – along with lots of human editing). The last six minutes of the talk include two demos on using SageMaker, CodePipeline, and CloudFormation as part of the open source solution we created.

Oracle has been named a leader in notebook-based Predictive Analytics and Machine Learning (PAML) solutions by Forrester Research, earning the highest average current offering score as well as the highest possible score for its solution roadmap. The Forrester Wave: Notebook-Based Predictive Analytics and Machine Learning Solutions, Q3 2018 report recognizes that Oracle Autonomous Data Science Cloud Service "provides the standardization and controls that enterprises need" and "makes it easy to put models into production by offering visual tools to create APIs with automatic load balancing." According to Forrester, PAML solutions are defined as "Software that provides enterprise data scientist teams and stakeholders with 1) tools to analyze data; 2) workbench tools to build predictive models using statistical and machine learning algorithms; 3) a platform to train, deploy, and manage analytical results and models; and 4) collaboration tools for extended enterprise teams including businesspeople, data engineers, application developers, DevOps, and AI engineers." Forrester evaluated the strengths and weaknesses of the top notebook-based PAML vendors across 24 evaluation criteria, which were grouped into three categories: current offering, strategy and market presence. Of the nine vendors Forrester evaluated, Oracle was one of the two companies recognized as a leader. "With Oracle Autonomous Data Science Cloud Service, Oracle has a winning solution for our customers to build and deploy artificial intelligence and machine learning models on the Oracle Cloud," said Greg Pavlik, Senior Vice President and Chief Technology Officer, Oracle Cloud Platform.

Those who learn how to make machines that exhibit intelligence today are tomorrow going to lead the next technological revolution, be part of the most cutting-edge companies and stand a chance to disrupt almost all industries through their skillsets.

In this blog we will try to get the basic idea behind reinforcement learning and understand what is a multi arm bandit problem. We will also be trying to maximise CTR(click through rate) for advertisements for a advertising agency. Article includes: 1. Basics of reinforcement learning 2. Types of problems in reinforcement learning 3. Understamding multi-arm bandit problem 4. Basics of conditional probability and Thompson sampling 5. Optimizing ads CTR using Thompson sampling in R Reinforcement Learning Basics Reinforcement learning refers to goal-oriented algorithms, which learn how to attain a complex objective (goal) or maximise along a particular dimension over many steps; for example, maximise the points won in a game over many moves. They can start from a blank slate, and under the right conditions, they achieve superhuman performance. Like a child incentivized by spankings and candy, these algorithms are penalized when they make the wrong decisions and rewarded when they make the right ones -- this is reinforcement.