Stability is a key aspect of data analysis. In many applications, the natural notion of stability is geometric, as illustrated for example in computer vision. Scattering transforms construct deep convolutional representations which are certified stable to input deformations. This stability to deformations can be interpreted as stability with respect to changes in the metric structure of the domain. In this work, we show that scattering transforms can be generalized to non-Euclidean domains using diffusion wavelets, while preserving a notion of stability with respect to metric changes in the domain, measured with diffusion maps. The resulting representation is stable to metric perturbations of the domain while being able to capture "high-frequency" information, akin to the Euclidean Scattering.

Gating is a key technique used for integrating information from multiple sources by long short-term memory (LSTM) models and has recently also been applied to other models such as the highway network. Although gating is powerful, it is rather expensive in terms of both computation and storage as each gating unit uses a separate full weight matrix. This issue can be severe since several gates can be used together in e.g. an LSTM cell. This paper proposes a semi-tied unit (STU) approach to solve this efficiency issue, which uses one shared weight matrix to replace those in all the units in the same layer. The approach is termed "semi-tied" since extra parameters are used to separately scale each of the shared output values. These extra scaling factors are associated with the network activation functions and result in the use of parameterised sigmoid, hyperbolic tangent, and rectified linear unit functions. Speech recognition experiments using British English multi-genre broadcast data showed that using STUs can reduce the calculation and storage cost by a factor of three for highway networks and four for LSTMs, while giving similar word error rates to the original models.

We present a first attempt to perform attentional word segmentation directly from the speech signal, with the final goal to automatically identify lexical units in a low-resource, unwritten language (UL). Our methodology assumes a pairing between recordings in the UL with translations in a well-resourced language. It uses Acoustic Unit Discovery (AUD) to convert speech into a sequence of pseudo-phones that is segmented using neural soft-alignments produced by a neural machine translation model. Evaluation uses an actual Bantu UL, Mboshi; comparisons to monolingual and bilingual baselines illustrate the potential of attentional word segmentation for language documentation.

This survey discusses how recent developments in multimodal processing facilitate conceptual grounding of language. We categorize the information flow in multimodal processing with respect to cognitive models of human information processing and analyze different methods for combining multimodal representations. Based on this methodological inventory, we discuss the benefit of multimodal grounding for a variety of language processing tasks and the challenges that arise. We particularly focus on multimodal grounding of verbs which play a crucial role for the compositional power of language.

Machine-learning has become one of the biggest buzzwords in cyber security, with almost every maker of a product in this sector touting it as part of their detection capability. A recent example of that turned up last week in a blog from Microsoft describing how in May Windows Defender spotted something suspicious in a small-scale email campaign purportedly from a landscaping business in Calgary. Microsoft's machine learning systems stopped the mail, which asked target victims to review an attached PDF document. It turned out the mail was coming from a spoofed address of the landscaping business. More than that, it was a spear phishing campaign to around 80 persons or firms.

The facial-recognition cameras installed near the bounce houses at the Warehouse, an after-school recreation center in Bloomington, Indiana, are aimed low enough to scan the face of every parent, teenager and toddler who walks in. The center's director, David Weil, learned earlier this year of the surveillance system from a church newsletter, and within six weeks he had bought his own, believing it promised a security breakthrough that was both affordable and cutting-edge. Since last month, the system has logged thousands of visitors' faces – alongside their names, phone numbers and other personal details – and checked them against a regularly updated blacklist of sex offenders and unwanted guests. The system's Israeli developer, Face-Six, also promotes it for use in prisons and drones. "Some parents still think it's kind of '1984,' " said Weil, whose 21-month-old granddaughter is among the scanned.

Machine learning is a key driver in the constant evolution of security technologies at Microsoft. Machine learning allows Microsoft 365 to scale next-gen protection capabilities and enhance cloud-based, real-time blocking of new and unknown threats. Just in the last few months, machine learning has helped us to protect hundreds of thousands of customers against ransomware, banking Trojan, and coin miner malware outbreaks. But how does machine learning stack up against social engineering attacks? Social engineering gives cybercriminals a way to get into systems and slip through defenses.

Models for audio source separation usually operate on the magnitude spectrum, which ignores phase information and makes separation performance dependant on hyper-parameters for the spectral front-end. Therefore, we investigate end-to-end source separation in the time-domain, which allows modelling phase information and avoids fixed spectral transformations. Due to high sampling rates for audio, employing a long temporal input context on the sample level is difficult, but required for high quality separation results because of long-range temporal correlations. In this context, we propose the Wave-U-Net, an adaptation of the U-Net to the one-dimensional time domain, which repeatedly resamples feature maps to compute and combine features at different time scales. We introduce further architectural improvements, including an output layer that enforces source additivity, an upsampling technique and a context-aware prediction framework to reduce output artifacts. Experiments for singing voice separation indicate that our architecture yields a performance comparable to a state-of-the-art spectrogram-based U-Net architecture, given the same data. Finally, we reveal a problem with outliers in the currently used SDR evaluation metrics and suggest reporting rank-based statistics to alleviate this problem.

NET ENERGY INC Founded in 2005, Net Energy Inc. is the original electronic trading platform and brokerage for crude oil in Canada. Net Energy connects motivated buyers and sellers in a transparent, reliable trading forum, and flows trade, settlement, and index data directly into customers' back-office risk, credit, and accounting applications. Headquartered in Calgary, Alberta, Canada, the knowledgeable and experienced trading desk is continually developing and expanding into new markets and currently offers over 300 physical energy and financial products available for transacting in real time. Net Energy is recognized by the Alberta Securities Commission as a Crude Oil Derivatives Exchange for its suite of financial products.

I'm driving down the San Diego freeway, searching for my exit, feeling jumpy and a little bit lost. I'm hoping a few wrong turns early in my drive won't make me late. I still need time to stop off at a strip mall with a Starbucks and a big parking lot -- someplace I can put on deodorant and maybe a little makeup after my cross-country flight. I prefer to think this is because I want to appear professional to the human man I'm on my way to see: Matt McMullen, the founder of Abyss Creations and its offshoot robotics company, Realbotix. But once I find a place to park and start brushing my hair, I realize that some deranged sliver of myself feels as if I'm primping for romance. This is a complicated realization. In addition to McMullen, I'm about to meet Henry, the first available male sex robot. Henry is six feet tall, with six-pack abs and the customer's choice of penis. He's just a prototype at the moment -- you can't buy him -- but the two female models Realbotix developed alongside Henry will ship this summer. So far, there have been 50 preorders at $12,000 apiece. Henry, Harmony, and Solana have sturdy silicone bodies, and once they're synced up to a corresponding app, they can give compliments, recite poetry, tell jokes, and seduce.