Roundup This week's AI roundup includes an alarming report from California's Department of Motor Vehicles about how shoddy autonomous cars still are, a Waymo self-driving car crash, and some news from Facebook's F8 conference and its new job posting. Uh oh, not another self-driving car crash It's Waymo's turn to be involved in a car crash. Reports from local news in Arizona showed a beat up Waymo van and a trashed Honda Sedan lying around piles of debris on a road in Chandler on Friday. Check out the impact from a TV report from ABC 15. The white Waymo van was in autonomous mode when the crash happened.

A machine learning learning PhD doesn't only open up some of the highest-paying jobs around, it sets you up to have an outsized positive impact on the world. This comprehensive guide on machine learning PhDs from 80,000 Hours (YC S15) will help you get started. The guide is based on discussion with six machine learning researchers including two at DeepMind, one at OpenAI, and one running a robotics start-up. Check out the highlights below. Machine learning involves giving software rules to learn from experience rather than directly programming the steps it takes.

When we think about the imminent development of the next digital revolution, humanity will face an unprecedented wave of automation. More and more smart and connected devices will coexist with us. This revolution is already taking place, from cell phones, to autonomous vehicles and even our refrigerator. Something is for sure, robots are already here, and they are here to stay. The question is not whether we agree, but how we will interact with these new tenants.

The answer to the question of what makes deep learning different from traditional machine learning for predictive... You forgot to provide an Email Address. This email address doesn't appear to be valid. This email address is already registered. You have exceeded the maximum character limit.

In natural language understanding, many challenges require learning relationships between two sequences for various tasks such as similarity, relatedness, paraphrasing and question matching. Some of these challenges are inherently closer in nature, hence the knowledge acquired from one task to another is easier acquired and adapted. However, transferring all knowledge might be undesired and can lead to sub-optimal results due to \textit{negative} transfer. Hence, this paper focuses on the transferability of both instances and parameters across natural language understanding tasks using an ensemble-based transfer learning method to circumvent such issues. The primary contribution of this paper is the combination of both \textit{Dropout} and \textit{Bagging} for improved transferability in neural networks, referred to as \textit{Dropping} herein. Secondly, we present a straightforward yet novel approach to incorporating source \textit{Dropping} Networks to a target task for few-shot learning that mitigates \textit{negative} transfer. This is achieved by using a decaying parameter chosen according to the slope changes of a smoothed spline error curve at sub-intervals during training. We compare the approach over the hard parameter sharing, soft parameter sharing and single-task learning to compare its effectiveness. The aforementioned adjustment leads to improved transfer learning performance and comparable results to the current state of the art only using few instances from the target task.

In this paper, benefiting from the strong ability of deep neural network in estimating non-linear functions, we propose a discriminative embedding function to be used as a feature extractor for clustering tasks. The trained embedding function transfers knowledge from the domain of a labeled set of morphologically-distinct images, known as classes, to a new domain within which new classes can potentially be isolated and identified. Our target application in this paper is the Gravity Spy Project, which is an effort to characterize transient, non-Gaussian noise present in data from the Advanced Laser Interferometer Gravitational-wave Observatory, or LIGO. Accumulating large, labeled sets of noise features and identifying of new classes of noise lead to a better understanding of their origin, which makes their removal from the data and/or detectors possible.

In this paper, we propose a self-attention mechanism, dubbed "fast directional self-attention (Fast-DiSA)", which is a fast and light extension of "directional self-attention (DiSA)". The proposed Fast-DiSA performs as expressively as the original DiSA but only uses much less computation time and memory, in which 1) both token2token and source2token dependencies are modeled by a joint compatibility function designed for a hybrid of both dot-product and multi-dim ways; 2) both multi-head and multi-dim attention combined with bi-directional temporal information captured by multiple positional masks are in consideration without heavy time and memory consumption appearing in the DiSA. The experiment results show that the proposed Fast-DiSA can achieve state-of-the-art performance as fast and memory-friendly as CNNs. The code for Fast-DiSA is released at \url{https://github.com/taoshen58/DiSAN/tree/master/Fast-DiSA}.

This paper presents a deep learning framework that is capable of solving partially observable locomotion tasks based on our novel Recurrent Deterministic Policy Gradient (RDPG). Three major improvements are applied in our RDPG based learning framework: asynchronized backup of interpolated temporal difference, initialisation of hidden state using past trajectory scanning, and injection of external experiences learned by other agents. The proposed learning framework was implemented to solve the Bipedal-Walker challenge in OpenAI's gym simulation environment where only partial state information is available. Our simulation study shows that the autonomous behaviors generated by the RDPG agent are highly adaptive to a variety of obstacles and enables the agent to traverse rugged terrains effectively.

Verifying correctness of deep neural networks (DNNs) is challenging. We study a generic reachability problem for feed-forward DNNs which, for a given set of inputs to the network and a Lipschitz-continuous function over its outputs, computes the lower and upper bound on the function values. Because the network and the function are Lipschitz continuous, all values in the interval between the lower and upper bound are reachable. We show how to obtain the safety verification problem, the output range analysis problem and a robustness measure by instantiating the reachability problem. We present a novel algorithm based on adaptive nested optimisation to solve the reachability problem. The technique has been implemented and evaluated on a range of DNNs, demonstrating its efficiency, scalability and ability to handle a broader class of networks than state-of-the-art verification approaches.

Non-synonymous single nucleotide polymorphisms (nsSNPs) are single nucleotide substitution occurring in the coding region of a gene and leads to a change in amino-acid sequence of protein. The studies have shown these variations may be associated with disease. Thus, investigating the effects of nsSNPs on protein function will give a greater insight on how nsSNPs can lead into disease. Breast cancer is the most common cancer among women causing highest cancer death every year. BRCA1 and BRCA2 tumor suppressor genes are two main candidates of which, mutations in them can increase the risk of developing breast cancer. For prediction and detection of the cancer one can use experimental or computational methods, but the experimental method is very costly and time consuming in comparison with the computational method. The computer and computational methods have been used for more than 30 years. Here we try to predict the clinical significance of BRCA1 and BRCA2 nsSNPs as well as the unknown clinical significances. Nearly 500 BRCA1 and BRCA2 nsSNPs with known clinical significances retrieved from NCBI database. Based on hydrophobicity or hydrophilicity and their role in proteins' second structure, they are divided into 6 groups, each assigned with scores. The data are prepared in the acceptable form to the automated prediction mechanisms, Probabilistic Neural Network (PNN) and Deep Neural NetworkStacked AutoEncoder (DNN). With Jackknife cross validation we show that the prediction accuracy achieved for BRCA1 and BRCA2 using PNN are 87.97% and 82.17% respectively, while 95.41% and 92.80% accuracies achieved using DNN. The total required processing time for the training and testing the PNN is 0.9 second and DNN requires about 7 hours of training and it can predict instantly. both methods show great improvement in accuracy and speed compared to previous attempts.