Datasets with significant proportions of noisy (incorrect) class labels present challenges for training accurate Deep Neural Networks (DNNs). We propose a new perspective for understanding DNN generalization for such datasets, by investigating the dimensionality of the deep representation subspace of training samples. We show that from a dimensionality perspective, DNNs exhibit quite distinctive learning styles when trained with clean labels versus when trained with a proportion of noisy labels. Based on this finding, we develop a new dimensionality-driven learning strategy, which monitors the dimensionality of subspaces during training and adapts the loss function accordingly. We empirically demonstrate that our approach is highly tolerant to significant proportions of noisy labels, and can effectively learn low-dimensional local subspaces that capture the data distribution.

In recent years, deep neural network is widely used in machine learning. The multi-class classification problem is a class of important problem in machine learning. However, in order to solve those types of multi-class classification problems effectively, the required network size should have hyper-linear growth with respect to the number of classes. Therefore, it is infeasible to solve the multi-class classification problem using deep neural network when the number of classes are huge. This paper presents a method, so called Label Mapping (LM), to solve this problem by decomposing the original classification problem to several smaller sub-problems which are solvable theoretically. Our method is an ensemble method like error-correcting output codes (ECOC), but it allows base learners to be multi-class classifiers with different number of class labels. We propose two design principles for LM, one is to maximize the number of base classifier which can separate two different classes, and the other is to keep all base learners to be independent as possible in order to reduce the redundant information. Based on these principles, two different LM algorithms are derived using number theory and information theory. Since each base learner can be trained independently, it is easy to scale our method into a large scale training system. Experiments show that our proposed method outperforms the standard one-hot encoding and ECOC significantly in terms of accuracy and model complexity.

Stochastic descent methods (of the gradient and mirror varieties) have become increasingly popular in optimization. In fact, it is now widely recognized that the success of deep learning is not only due to the special deep architecture of the models, but also due to the behavior of the stochastic descent methods used, which play a key role in reaching "good" solutions that generalize well to unseen data. In an attempt to shed some light on why this is the case, we revisit some minimax properties of stochastic gradient descent (SGD) on the square loss of linear models---originally developed in the 1990's---and extend them to generic stochastic mirror descent (SMD) algorithms on general loss functions and nonlinear models. In particular, we show that there is a fundamental identity which holds for SMD (and SGD) under very general conditions, and that this identity implies the minimax optimality of SMD (and SGD) with sufficiently small step size, for a general class of loss functions and general nonlinear models. We further show that this identity can be used to naturally establish other properties of SMD (and SGD), such as convergence and "implicit regularization" for over-parameterized linear models, which have been shown in certain cases in the literature. We also show how this identity can be used in the so-called "highly over-parameterized" nonlinear setting to provide insights into why SMD (and SGD) may have similar convergence and implicit regularization properties for deep learning.

Recommender systems play a crucial role in mitigating the problem of information overload by suggesting users' personalized items or services. The vast majority of traditional recommender systems consider the recommendation procedure as a static process and make recommendations following a fixed strategy. In this paper, we propose a novel recommender system with the capability of continuously improving its strategies during the interactions with users. We model the sequential interactions between users and a recommender system as a Markov Decision Process (MDP) and leverage Reinforcement Learning (RL) to automatically learn the optimal strategies via recommending trial-and-error items and receiving reinforcements of these items from users' feedback. Users' feedback can be positive and negative and both types of feedback have great potentials to boost recommendations. However, the number of negative feedback is much larger than that of positive one; thus incorporating them simultaneously is challenging since positive feedback could be buried by negative one. In this paper, we develop a novel approach to incorporate them into the proposed deep recommender system (DEERS) framework. The experimental results based on real-world e-commerce data demonstrate the effectiveness of the proposed framework. Further experiments have been conducted to understand the importance of both positive and negative feedback in recommendations.

According to Gartner's hype cycle of emerging technologies, 2017; Deep Learning and Machine Learning have reached the peak of inflated expectations. Artificial General Intelligence (AGI) and Deep Reinforcement Learning are in the phase of innovation trigger. The sentiment over Artificial Intelligence (AI) is euphoric. Every technology firm is jumping on the AI first bandwagon. Companies like Google, Microsoft, Amazon, and Alibaba are pushing the frontiers.

Artificial intelligence is an important topic in the gaming industry, whether it is a sports game, first-person shooter or an RPG. It is a constantly evolving aspect of gaming, and Hellblade: Senua's Sacrifice director Tameem Antoniades is interested in its development. Antoniades expressed his interest in AI and what can "machine learning and deep learning" achieve over the next few years. I am interested in AI, because finally we're breaking through. AI technology has basically been in the doldrums for 30 or 40 years with very little in the way of advancement, and finally we're getting really good results – eye-opening results.

Artificial intelligence is revolutionizing our world in many unimaginable ways. At the verge of the Fourth Industrial Revolution, humanity is currently witnessing the first steps made by machines in reinventing the world we live in. And while we keep debating about the potential drawbacks and benefits of substituting humans with intelligent, self-learning machines, there's one area where AI's positive impact will definitely improve the quality of our lives: the health care industry. Machine learning algorithms can process unimaginable amounts of info in the blink of an eye. And they can be much more precise than humans in spotting even the smallest detail in medical imaging reports such as mammograms and CT scans.

Taraneh Khazaei is currently a data scientist at Microsoft, helping Microsoft clients from a wide variety of industries (e.g., finance, insurance, and retail) solve, build, and deploy their AI solutions. She holds a Ph.D. degree in computer science, where she focused on a variety of machine learning and statistical modeling techniques in the context of natural language processing, social computing, and online privacy mining. Before joining Microsoft, she worked as a data scientist at Zero Gravity Labs (ZGL), researching the state of the art of deep learning. Prior to ZGL, Taraneh worked at TD Bank as a data scientist/engineer and at InfoTrellis as a machine learning researcher.

Thanks to all the wonderful folks at DataScience.com and especially Pramit Choudhary for helping me discover the amazing world of model interpretation. The field of Machine Learning has gone through some phenomenal changes over the last decade. Starting off as just a pure academic and research-oriented domain, we have seen widespread industry adoption across diverse domains including retail, technology, healthcare, science and many more. Rather than just running lab experiments to publish a research paper, the key objective of data science and machine learning in the 21st century has changed to tackling and solving real-world problems, automating complex tasks and making our life easier and better. More than often, the standard toolbox of machine learning, statistical or deep learning models remain the same.

In this article, let's check about some of the best frameworks and libraries for Machine Learning. This list is created by me based on a variety of parameters, some would surely not accept it but again it is according to me and would vary from person to person. If you are a beginner, check out our articles on "Machine learning crash course" and "Machine learning specialization course". Each of these Frameworks is different from each other and takes much time to learn, during the time of making this list we took care of features other than the basic ones, User base and community & support was one of the most important parameters. Some frameworks are more mathematically oriented, and hence geared more towards statistical than neural networks. Some of them provide a rich set of linear algebra tools; some are mainly focused only on deep learning.