Author: Smit Mukeshkumar Patel

As we see the upward trend of daily cases in COVID-19 there is also a proportional upward trend in the overload of paperwork in the health care system. More people are needed to be kept track of and hence there is a exponential amount of growth in available data. Once we reach a certain threshold of data gathered, it simply is not possible for humans to use all of the data to perform analysis on top of. Hence, the usage of artificial intelligence and computers become essential to perform heavy analysis. Maithra Raghu, PhD student at Cornell CS and Google Brain has been doing extensive research into this matter and wants to leave the the focus on data to computer scientists so doctors can focus on their patients instead.

Raghu has proposed to use neural networks to perform analysis on vast amounts of information and help health experts find patterns that human eye cannot pick up. We can further use these patterns to help diagnose and predict much earlier than if we were to use human effort. More importantly we can have more hands helping patients fighting the disease than having them go through paperwork that can be taken care of by computers.

Using these amazing technologies such as neural networks to help solve COVID-19 is only one of the great example of how powerful and useful network analysis and neural networks are. Google Brain was able to use large amounts of data on brain to create a 3D model that traces 20 million synapses connecting more than 25,000 neurons in the brain of a fruit fly. It is estimated that human beings only use 10% of their brain’s capacity, with the advancements of neural networks and network analysis, I believe we will be able to better understand the human brain and access higher usage of the brain.

In conclusion, I believe the work on neural networks is really interesting and is being used mostly everywhere in the tech industry (Apple Silicon). I also believe it will play a big role in data analysis and help us solve problems which are simply unsolvable with human effort. Computers will help us become smarter and reduce tasks which it can do it better and faster than us.

References

Chen, J. (2020, November 13). A Google Brain scientist turns to AI to make medicine more personal. Retrieved November 20, 2020, from https://www.statnews.com/2020/11/16/google-brain-maithra-raghu-artificial-intelligence/

Vincent, J. (2020, January 22). Google publishes largest ever high-resolution map of brain connectivity. Retrieved November 20, 2020, from https://www.theverge.com/2020/1/22/21076806/google-janelia-flyem-fruit-fly-brain-map-hemibrain-connectome

When it comes to performing analysis on unstructured data we usually forget how much data there is to perform computation on and how much it relies on scale computing. The terms like Big Data, Map-Reduce, and Parallel programming get thrown around commonly nowadays and many engineers are working towards building computer architecture which will help enable efficient computation of unstructured data and tackle large-scale problems. Intel and Katana Graph Team have started a collaboration to port and optimize the Katana Graph Engine on Intel Xeon for scalable processing. What this will do is enable users to exploit high-performance, scale-out parallel computing and compute large problems with unstructured data such as network analysis with unmatched efficiency. In the modern era unstructured datasets are very common in social network analysis, security and authentication, biomedical and pharma applications and many more.

One key usage of large unstructured data sets and performing network analysis that we all are aware of by now is epidemiological studies for modeling the spread of infectious diseases like COVID-19. The Katana Graph Engine performs computation on any problem involving connected data and hence all network analysis will be enabled through this upgrade. The Katana Graph Engine has been widely used as a major defense contractor to implement a system that will detect real-time intrusion in computer network(s) and performs heavy data mining on how users interact in the network and analyze their patterns. All of these tasks take long time if you consider the number of users to be in billions and which is why concepts such as Parallel Programming and Big Data exists where we enable the use of extra hardware to increase efficiency.

The collaboration of Intel and Katana Graph Engine is just one example of how many companies are looking ahead into the future and see the need for heavy network analysis on unstructured data. Many are starting to adapt to idea of using such technologies to gain better information through Machine Learning and Artificial Intelligence to improve their software. Through network analysis and quick network analysis it will enable us to discover interrelations between sets of metrics which is not possible today and there are layers upon layers of insights that can be derived from a single data point. There is data hidden inside data and there are unique connections between these layers depending on how one perceives it. Therefore I believe this is an exciting collaboration of hardware and software which could revolutionize how we perceive network analysis the extent to which it can be performed.

References

Group, K., 2020. Katana Graph Collaborating With Intel On Enterprise-Scale Computing On Unstructured Data :: Katana Graph. [Accessed 22 October 2020] https://katanagraph.com/news/katana-graph-intel-collaboration

H. (2020, October 19). Intel and Katana Graph Team on Large-scale Graph Analytics. Retrieved October 22, 2020, from https://www.hpcwire.com/off-the-wire/intel-and-katana-graph-team-on-large-scale-graph-analytics/