Earlier, we came up with a list of some of the best Machine Learning books that you should consider reading through. In this article, we have come up with yet another list of the recommended books for Data Science. Written by Blum, Hopcroft, and Kannan, Foundations of Data Science is a great blend of lectures in the modern theoretical course in data science. This tutorial on UFLDL aims to get you familiar with the main ideas of Unsupervised Feature Learning and Deep Learning. The Python Data Science Handbook introduces the core libraries essential for working with data in Python -- particularly IPython, NumPy, Pandas, Matplotlib, Scikit-Learn, and related packages.

Enjoy it while it lasts. US Treasury Secretary Steven Mnuchin may think artificial intelligence (AI) isn't going to start taking humans' jobs for 50 to 100 years, but most experts believe a revolution in automation is coming far sooner, promising massive increases in efficiency -- and job losses on a huge scale. A recent study put out by PwC estimated that as many as 30% of UK jobs could be "susceptible to automation by robots and AI" by the early 2030s -- with 38% in the US at risk, 35% in Germany, and 21% in Japan -- although it believes jobs will be created elsewhere in the economy to help offset this. Are we doing enough to prepare? Absolutely not, says Bob Moritz, global chairman of consultancy firm PwC.

China has great potential in applications of artificial intelligence (AI), a senior official said Sunday. "Chinese researchers and entrepreneurs are among the best in the world, with technological innovations and good earnings in the sector," said Liu Lihua, vice minister of industry and information technology. Researchers with Chinese companies such as iFlytek, Alibaba and Baidu participated in the study of the world's leading AI technologies, said Liu, referring to technologies of reinforcement learning, paying with your face and self-driving trucks. A couple of weeks ago, the "MIT Technology Review" listed the above three and another seven technologies as its 10 breakthrough technologies in 2017. AI research started more than 60 years ago and there have been some major ups and downs.

Newswire) Digital Reasoning, a leader in cognitive computing, was announced as the company with the'Best Artificial Intelligence Technology' at the WatersTechnology.com Digital Reasoning created the Synthesys cognitive computing platform, which is one of the most widely-adopted AI systems within financial services and used by many of the world's leading investment banks in applications ranging across risk and compliance, financial crime, and customer insights. Synthesys understands and analyzes human communications. Most communications data is unstructured, making it virtually unreadable using conventional technology. By applying artificial intelligence, Synthesys makes sense of human language in text, audio and images, and resolves who is talking about what and with whom.

Newswire) Remesh, a software company that is reinventing market research through artificial intelligence (AI), today announced the closing of its $2.25 million seed investment round that brings its total funding to $3.85 million. The round is led by LionBird Ventures, a venture capital firm investing in early stage digital health and business services companies with offices in Tel Aviv and Chicago. The round also includes Reimagine Holdings Group, a holding company focused on growing consumer insights and marketing services companies, as well as individual investors, representing a mix of new and returning investors. "We believe that Remesh has shown real potential to change the way brands, consultants and agencies listen to feedback from their audiences," said Ed Michael, Managing Partner at LionBird Ventures. "Remesh has recognized a way to solve for a number of inefficiencies in market research using artificial intelligence. This new model not only replaces legacy systems, but establishes entirely new market research workflows."

Facebook's AI assistant, known simply as M, will now pop into your Messenger chat windows to suggest actions it can take on your behalf, the company announced today. The feature is rolling out to iOS and Android users in the US, with a broader expansion around the globe in the coming months. Facebook first began testing this feature in December, and it appears ready to be unleashed on the public. The current system works by analyzing your conversation and looking for key words to trigger M's suggestive capabilities. Those capabilities include sending stickers on your behalf, initiating payment requests through Messenger, calling a ride-hailing app like Uber and Lyft, starting a poll for group chat participants, and sharing your location with others. M will also look for key words that suggest you're trying to make plans with a friend and jump in to help coordinate that.

New Delhi, Most customers are confused about the use of artificial intelligence (AI) and are, therefore, reluctant to embrace this new technology, a study said on Friday. Released by US-based software firm Pegasystems, it revealed that these fears are often eased once the users gain firsthand AI experience -- which ironically many enjoy without even realising it. "Our study suggests the recent hype is causing some confusion and fear among consumers, who may not really understand how it's already being used and helping them every day," said Don Schuerman, Vice President (Product Marketing) Pegasystems. The study that involved 6,000 customers in six countries found that consumers were hesitant to fully embrace AI devices and services. "Only 36 per cent are comfortable with businesses using AI to engage with them. Almost 72 per cent express some sort of fear about AI," the study found.

What makes images similar? To measure the similarity between images, they are typically embedded in a feature-vector space, in which their distance preserve the relative dissimilarity. However, when learning such similarity embeddings the simplifying assumption is commonly made that images are only compared to one unique measure of similarity. A main reason for this is that contradicting notions of similarities cannot be captured in a single space. To address this shortcoming, we propose Conditional Similarity Networks (CSNs) that learn embeddings differentiated into semantically distinct subspaces that capture the different notions of similarities. CSNs jointly learn a disentangled embedding where features for different similarities are encoded in separate dimensions as well as masks that select and reweight relevant dimensions to induce a subspace that encodes a specific similarity notion. We show that our approach learns interpretable image representations with visually relevant semantic subspaces. Further, when evaluating on triplet questions from multiple similarity notions our model even outperforms the accuracy obtained by training individual specialized networks for each notion separately.

We introduce a novel loss max-pooling concept for handling imbalanced training data distributions, applicable as alternative loss layer in the context of deep neural networks for semantic image segmentation. Most real-world semantic segmentation datasets exhibit long tail distributions with few object categories comprising the majority of data and consequently biasing the classifiers towards them. Our method adaptively re-weights the contributions of each pixel based on their observed losses, targeting under-performing classification results as often encountered for under-represented object classes. Our approach goes beyond conventional cost-sensitive learning attempts through adaptive considerations that allow us to indirectly address both, inter- and intra-class imbalances. We provide a theoretical justification of our approach, complementary to experimental analyses on benchmark datasets. In our experiments on the Cityscapes and Pascal VOC 2012 segmentation datasets we find consistently improved results, demonstrating the efficacy of our approach.

Empirical risk minimization (ERM) is ubiquitous in machine learning and underlies most supervised learning methods. While there has been a large body of work on algorithms for various ERM problems, the exact computational complexity of ERM is still not understood. We address this issue for multiple popular ERM problems including kernel SVMs, kernel ridge regression, and training the final layer of a neural network. In particular, we give conditional hardness results for these problems based on complexity-theoretic assumptions such as the Strong Exponential Time Hypothesis. Under these assumptions, we show that there are no algorithms that solve the aforementioned ERM problems to high accuracy in sub-quadratic time. We also give similar hardness results for computing the gradient of the empirical loss, which is the main computational burden in many non-convex learning tasks.