Regularization does NOT improve the performance on the data set that the algorithm used to learn the model parameters (feature weights). However, it can improve the generalization performance, i.e., the performance on new, unseen data, which is exactly what we want. In intuitive terms, we can think of regularization as a penalty against complexity. Increasing the regularization strength penalizes "large" weight coefficients -- our goal is to prevent that our model picks up "peculiarities," "noise," or "imagines a pattern where there is none." Again, we don't want the model to memorize the training dataset, we want a model that generalizes well to new, unseen data. In more specific terms, we can think of regularization as adding (or increasing the) bias if our model suffers from (high) variance (i.e., it overfits the training data).

Today in "terrifying things you can't look away from" is smilevector, a Twitter bot that uses a neural network to take photos from the internet and add or remove smiles. Smilevector was created this month by Tom White, a lecturer at the Victoria University of Wellington School of Design in Wellington, New Zealand. The bot is based on generative neural networks, which means it was trained by looking at lots of smiles, and has now learned to add or remove them to other photos. You must obey your smile overlord.

Who do we become when we die? Our identities are such fragile, personally curated things: we go through life as children, then students and peers, and sometimes switch to a vocational preference. A person becomes farmer, shepherd, mechanic, soldier, baker, homemaker, brigand, or bandit, or backyard bomb-maker -- whichever strikes us by calling or necessity. People assume narrower identities, a lover for an afternoon, a wedding guest for a weekend. Perhaps it's the wrong wedding, the wrong place, the wrong people, and whatever mish-mash of identities, they end with a hellfire strike and a grim, clinical finality. The bodies become "military-age males," the rich matrixes of interwoven identities collapsed into two categories, a fatal guilt decided abroad in the moment of impact.

Today on In Case You Missed It: Boston Dynamics introduced a new robot to the lineup and this one is both the smallest yet, and most adorable. The SpotMini is just two feet tall but in the video Boston Dynamics released, shows it can do dishes, throw cans away and creep around for a good 90 minutes on an electric charge. Meanwhile the EPFL designed a new medical device that looks not-entirely unlike an old school slap bracelet, except it's made of silicon and is designed to grip and squeeze the aorta, keeping the heart beating while a patient might be waiting for a heart transplant. Theater fans will want to know about New York's decision to ban robot buyers; while politically-minded folks (not already talking about Great Britain's vote to leave the EU) will want to hear about C-SPAN's decision to broadcast Periscope videos of the Democrat's sit-in. As always, please share any interesting tech or science videos you find by using the #ICYMI hashtag on Twitter for @mskerryd.

Self-driving cars may be the new direction for the automotive industry, even though realistically it could be decades before we start seeing them all around us. That isn't stopping some from pondering a moral dilemma these cars may pose. When you take the control away from the driver, you are giving the motor vehicle power to make judgment calls. In many ways, you give a machine the power to choose your fate in specific scenarios. This is the very thing physicist Stephen Hawking warned us about, as he believes artificial intelligence could lead to a very real Terminator scenario.

A new computer prototype called a "memcomputer" works by mimicking the human brain, and could one day perform notoriously complex tasks like breaking codes, scientists say. These new, brain-inspired computing devices also could help neuroscientists better understand the workings of the human brain, researchers say. In a conventional microchip, the processor, which executes computations, and the memory, which stores data, are separate components. In contrast, Massimiliano Di Ventra, a theoretical physicist at the University of California, San Diego, and his colleagues are building "memcomputers," made up of "memprocessors," that both process and store data. This setup mimics the neurons that make up the human brain, with each neuron serving as both the processor and the memory. The building blocks of memcomputers were first theoretically predicted in the 1970s, but they were manufactured for the first time in 2008.

Artificial intelligence has crept into our everyday lives, even though we're not always aware of it. Here are some examples of AI's many applications. Journalists might not care for this, but AI programs are becoming smart enough to compile bits of information and turn them into articles. What's critical to understand about the potential impact machine learning can have on a manufacturing business lies in the way data is studied. The use of artificial intelligence that relies on natural language understanding is vital for achieving the potential outcomes mentioned above - otherwise, market intelligence would be based purely on statistical patterns and not contextual meaning.

In what is being perceived as another attempt to tighten its control over the internet, China's internet regulator on Saturday announced new rules that ban search engines from showing subversive information and obligate them to clearly identify paid results. The new regulations, which will take effect from Aug. 1, come close on the heels of the death of a 21-year-old college student, who is believed to have undergone an unapproved, experimental cancer treatment he found using the search engine Baidu. "Some search results lack objectivity and fairness, go against corporate morals and standards, misleading and influencing people's judgment," the Cyberspace Administration of China -- the country's internet regulator -- reportedly said. "Internet search providers should earnestly accept corporate responsibility toward society, and strengthen their own management in accordance with the law and rules, to provide objective, fair and authoritative search results to users." In addition, search engines would also be required to censor "rumors, obscenities, pornography, violence, murder, terrorism and other illegal information" -- regulations that the Chinese government claims are needed to safeguard the security of its citizens.

Gensim is deep learning toolkit implemented in python programming language intended for handling large text collections, using efficient algorithms.

As the automotive ecosystem pushes forward with evermore-connected vehicles, new solutions – from over-the-air software updates to connected car services – are emerging from vendors to manage the vast amounts of data that accompany that connectivity. Airbiquity recently announced a software and data management solution specifically for the automotive industry to manage large-scale data collection and software updates to vehicles. SAP introduced its cloud-based Vehicle Insights tool that leverages its HANA big data platform to analyze telematics information as well as other existing business data and external data, with the aim of better integrating connected vehicles into business processes. Last week, IBM said it will be partnering with Local Motors of Maryland on a small, autonomous bus named Olli that incorporates IBM's artificial intelligence and analytics software, Watson, in its first vehicle-based "internet of things" solution – with the big data capabilities focused on passenger interactions rather than the vehicle's self-driving capabilities. Gartner has predicted that by 2020, connected car services will generate almost 40 billion in revenue annually, driven by a combination of infotainmet, navigation, fleet management, traffic management, remote diagnostics and automotive crash notification, among others.