Full duplex may be the next breakthrough in mobile networking

Stanford startup Kumu Networks didn’t receive much notice at Mobile World Congress this week as the giants of the mobile industry revealed their plans for 2015, but it did get the attention of two rather important mobile carriers. At their separate booths, Telefónica and SK Telecom were showing off a Kumu-built radio transmission system called full duplex, which both carriers said could eventually become one of the key technologies of any future 5G standard.

When the mobile companies pull out the 5G card, they’re usually trying to signal that something is a really big deal, and in the case of Kumu, they could very well be right. What full duplex does is solve a fundamental problem in wireless communications that limits a network’s full capacity potential: the inability to transmit and receive signals on a radio channel at the same time. The problem is known as self-interference, but the concept is not quite as complex as it sounds.


Imagine two people are having a conversation, which itself is one of the simplest two-way — or duplex — communication channels. If both people are talking at the same time, neither one can understand what the other is saying. The words one person speaks get drowned out by the other’s voice before it ever reaches his ears. The same principle holds for wireless transmissions. When a radio is transmitting its signals bleed over into its own receiver interfering with the signals it’s trying to listen for.

For that reason wireless networks have always been built in something called half-duplex mode, which basically prevents them from ever transmitting and receiving in the same channel at the same time. It’s why most mobile networks in the world today use different sets of frequencies for downlink and uplink transmissions (For instance in many U.S. LTE systems, our devices receive data from the tower in a 2100 MHz channel, but they send information back at 1700 MHz). And it’s why a Wi-Fi router flip-flops between transmitting and receiving when it talks to your laptop or smartphone. Half-duplex has served the wireless industry well, but using it means you’re only using half of the total capacity of your airwaves at any given time.

Kumu Networks is based in Santa Clara but its roots are in Stanford where its founders started their full duplex research.

Kumu Networks is based in Santa Clara but its roots are in Stanford where its founders started their full duplex research.


As my colleague Signe Brewster wrote in Gigaom’s first look at the Stanford startup in 2013, Kumu claims to have developed the mathematical breakthrough necessary to solve the problem of self-interference at a practical level. And now it’s claiming to have produced a commercially viable full-duplex radio system that can transmit and receive simultaneously without turning its connection to mush. According to Kumu VP of product development Joel Brand, the company accomplished this by becoming a very smart listener.

Essentially Kumu is constantly scanning the radio environment, gauging the exact state of the airwaves at any given time, Brand said. Using internally developed algorithms, Kumu can “hear” how the transmission the radio is pumping out is changing the signal environment a the receiver. It can then compensate for those changes as signals heading the opposite direction arrive. It’s like echo cancellation applied to radio waves instead of sound.

Full Duplex demo

Kumu supplied some photos of the full duplex rig it demoed at Mobile World Congress, and I’ll be the first to admit it doesn’t look very impressive. But at MWC I asked Vish Nandlall, CTO of Australian multinational mobile carrier [company]Telstra[/company], about the technology, and he said it was the real deal. Full duplex isn’t some crazy new concept Kumu just made up one day, he said. Full duplex is used today in regular phone lines, and its application to wireless has been kicking around scientific papers and academic research labs for some time. But what Kumu did was come up with a viable technology that could be applied to real world networks, Nandlall said.

The impact could be quite significant. If you remove the self-interference barrier, carriers could use all of their spectrum for both uplink and downlink at the same time, which would double the capacity or double the number of connections any network could support. Wi-Fi networks would no longer have to alternate between sending data and receiving it, thus dramatically improving their download and upload speeds. It might not solve the so-called spectrum crunch, but it would go a long way to making wireless networks a lot more efficient.

Right now Kumu is pitching the technology to carriers as a backhaul system, so they could use their 4G spectrum to concurrently communicate with phones and the core network. But Brand says in the future full duplex can easily be applied to the access network connecting our devices. In fact, Kumu’s MWC demos were using off-the-shelf radio smartphone chips from [company]Qualcomm[/company], just with the duplexer ripped out. That kind of change would require a redesign of both our networks and our devices, which isn’t going to happen overnight. That’s why Kumu and its carrier partners [company]Telefónica[/company] and [company]SK Telecom[/company] are looking ahead to 5G.


Google, Stanford say big data is key to deep learning for drug discovery

A team of researchers from Stanford University and Google have released a paper highlighting a deep learning approach they say shows promise in the field of drug discovery. What they found, essentially, is that that more data covering more biological processes seems like a good recipe for uncovering new drugs.

Importantly, the paper doesn’t claim a major breakthrough that will revolutionize the pharmaceutical industry today. It simply shows that by analyzing a whole lot of data across a whole lot of different target processes — in this case, 37.8 million data points across 259 tasks — seems to work measurably better for discovering possible drugs than does analyzing smaller datasets and/or building models specifically targeting a single a task. (Read the Google blog post for a higher-level, but still very-condensed explanation.)

But when talking about a process in drug discovery that can take years and cost drug companies billions of dollars that ultimately make their way into the prices of prescription drugs, any small improvement helps.

This graph shows a measure of prediction accuracy (ROC AUC is the area under the receiver operating characteristic curve) for virtual screening on a fixed set of 10 biological processes as more datasets are added.

This graph shows a measure of prediction accuracy (ROC AUC is the area under the receiver operating characteristic curve) for virtual screening on a fixed set of 10 biological processes as more datasets are added.

Here’s how the researchers explain the reality, and the promise, of their work in the paper:

The efficacy of multitask learning is directly related to the availability of relevant data. Hence, obtaining greater amounts of data is of critical importance for improving the state of the art. Major pharmaceutical companies possess vast private stores of experimental measurements; our work provides a strong argument that increased data sharing could result in benefits for all.

More data will maximize the benefits achievable using current architectures, but in order for algorithmic progress to occur, it must be possible to judge the performance of proposed models against previous work. It is disappointing to note that all published applications of deep learning to virtual screening (that we are aware of) use distinct datasets that are not directly comparable. It remains to future research to establish standard datasets and performance metrics for this field.

. . .

Although deep learning offers interesting possibilities for virtual screening, the full drug discovery process remains immensely complicated. Can deep learning—coupled with large amounts of experimental data—trigger a revolution in this field? Considering the transformational effect that these methods have had on other fields, we are optimistic about the future.

If they’re right, we might look back on this research as part of a handful of efforts that helped spur an artificial intelligence revolution in the health care space. Aside from other research in the field, there are multiple startups, including Butterfly Network and Enlitic (which will be presenting at our Structure Data conference later this month in New York) trying to improve doctors’ ability to diagnose diseases using deep learning. Related efforts include the work IBM is doing with its Watson technology to analyze everything from cancer to PTSD, as well as from startups like Ayasdi and Lumiata.

There’s no reason that researchers have to stop here, either. Deep learning has proven remarkably good at tackling machine perception tasks such as computer vision and speech recognition, but the approach can technically excel at more general problems involving pattern recognition and feature selection. Given the right datasets, we could soon see deep learning networks identifying environmental factors and other root causes of disease that would help public health officials address certain issues so doctors don’t have to.

Vivek Wadhwa steps back from the women-in-tech debate

Vivek Wadhwa, the academic and researcher who has been a vocal critic of high-tech’s male-dominated culture and a proponent of women’s role in technology,  is stepping back from that fight, according to a post he wrote for the Washington Post.

Wadhwa, who is affiliated with Duke and Stanford, as well as the for-profit Singularity University, has faced criticism — which he denies — that he profited from his advocacy of women in technology. What appears to have been the last straw was a WNYC TLDR podcast, subsequently removed, featuring criticism of Wadhwa by Amelia Greenhall, co-founder of Model View Culture.

Wadhwa complained that he was given no chance to respond. He ended up providing that reaction in a Huffington Post blog in which he pointed readers to a cached version of the podcast, so that listeners can make their own decisions. WNYC did air  Wadhwa’s response to the initial podcast.

A big part of the issue critics have is that they don’t think that men are qualified to speak on the topic of women in technology. Women technologists don’t need some guy to “mansplain” their issues, is the gist. The Financial Times has more on that here.

Note: This story was updated at 9:39 a.m. PST to add a link to WNYC podcast with Wadhwa’s response. 

Closing the academia-startup gap

The high-tech industry, heck industry in general, would be better off if academic researchers could bring the fruits of their labor to market faster. That’s an old argument, brought up anew in a blog by Matt Welsh, a software engineer at Google.

Battery startup Prieto charges up with funds

A lithium ion battery that can charge in five minutes and last for five times longer than the standard — that’s the goal for startup Prieto Battery, which just raised $5.5 million of a planned $6.8 million round.