Hello everyone,
I would like to propose a discussion about possibilities offered by Deep Learning in fighting epidemics. This week I would like to present the following article: Malaria Likelihood Prediction By Effectively Surveying Households Using Deep Reinforcement Learning. This is an example of a quite simple application of an algorithm which benefited in a real life.
Authors built a deep reinforcement learning (RL) agent which is capable of predicting the likelihood of an individual testing positive for malaria by questioning about their household.
The agent learnt which survey question should be asked next and when to stop to make a prediction.
It was possible to make a prediction with a high accuracy. Authors proved that adaptive survey can significantly reduce the number of questions needed to attain the high accuracy. As a result a large-scale and cost-effective monitoring of malaria was enabled through SMS surveys.
My questions:
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteI have no idea about any examples.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
Well, as I mentioned I am not a specialist, but it seems to me that it might not be enough. As far as I know, the greatest threat are people who constantly travel to the countries of the third world and carry diseases personally. It is not difficult, because just one person who is on the plane will infect other passengers and they will infect other people. In this way an epidemic can be created. It is also difficult to stop this process, because you would have to have an odious government ban on entering a country, which is unlikely.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
With the explosion of social media sites and proliferation of digital computing devices and Internet access, massive amounts of public data is being generated on a daily basis. Efficient techniques/ algorithms to analyse this massive amount of data can provide near real-time information about emerging trends and provide early warning in case of an imminent emergency (such as the outbreak of a viral disease).
Thank you for the response. Undoubtedly, deep learning may support decision-makers. It is desirable to indicate which are the most vulnerable regions in the context of population movements.
Delete
ReplyDeleteHmm I don't think that doing surveys in Africa using SMS is good idea, this is similar aproach as it was in past with US president election and phones.
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
Sorry but I don't know any other examples of this kind application.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I think yes because biggest adventage of Deep Learning is that, that can analyzie much more futures than human can do. With this it can give us prediction more accurate and also it can learn from data from other continents and countries.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
Big plus of deep learning is method called Online Learnig which can gives you predictions in real time and it is also learning constant. So it can analyze whap is happening now and predict next steps. It can be used with all ANN which are available.
Hi, thanks for the answers and mentioning the Online Learning approach. Online Learning seems to be promising because in fighting such epidemics we are mainly interested not in explaining the past determinants but in indicating the next steps.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteThanks for this artcile. It is good to read about new applications of Deep learning. Unfortuatelly this is the first time I hear about fighting epidemic with deep learning so I cant give any other example.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I think yes, we are able to. It is common to make models of such epidemic events using different tools so I am sure artifficial neural networks are able to learn that.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
For sure the most popular approaches like LSTM, CNN or auto-encoder but it is hard to say without previous deep investigation.
Hi, thank you for the response. Regarding the third question I agree with you - it all depends on the particular task - i.e. what is the object we want to investigate.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteFirst of all, thank you very much for an interesting article. I really enjoyed it. Unfortunately I don't know any other examples of Deep Learning research applications used for fighting epidemics, but I will definitely try to explore this subject in the nearest future.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I agree with Cezary. With so many people travelling between countries and continents back and forth we're not really able to foresee and prevent epidemics. Of course we can predict some of them using Deep Learning, but definitely not all of them.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
I'm not an expert in this field, so it's hard for me to say.
Hi, thanks for the comments. The increased mobility of individuals, even to the distant and exotic destinations is a challenge.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteI don't know any similar examples.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I think a good usage of Deep Learning can help with many things. I'm not convinced to the way presented here, but systems taking into account other factors, being able to analyze some medical data or to get some information about the health after chatting, should help preventing the outbreaks.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
As I mentioned, more date should be taken into account. I agree that information form social media can be used here as well as a location data. But this is just a tip of the mountain and I'm not the expert here.
Hi, thank you for the answers. It is important to be able to process as much information as it is possible. It is also vital to indicate which data has an impact in order reduce the time spent on analysis.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeletePerhaps it is not real 'deep learning', but social media uses something similar in case of depression. After surveys and research - there was prepared a list of common messages/sentences and even photo images that suggest depressed people. As a user, when you start to use similar words/posting similar photos (similar to the researched pattern), you will be informed that "hey, perhaps you need help?" and also you will see content that probably will help you to understand the problem and how to get help.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
Yes, as mentioned above, if research is properly done, you can use deep learning and prevent some problems from something happen.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
Perhaps the research with live updates on the status. Research including for example data from hospitals (what disease) merged with the place of living of sick people.
Hello, thanks for the comments. I have also heard about the new facility of the depression detection. Unfortunately, I don't think I know what kind of algorithms it uses. Facebooks AI will also detect a potential suicide by the analysis of the recent posts. Apparently, It is banned in UE - https://businessinsider.com.pl/technologie/nowe-technologie/facebook-rozpoznaje-potencjalnych-samobojcow/98nslyf .
DeleteThanks for the article - it is a good opportunity to learn something new about DL. I
ReplyDelete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
I almost thought that I have no idea about DL being used for that. Classic ML definitely - running medical records to aid the diagnosis of a doctor etc. But DL? Problably not yet?
Well, one example may be the segmentation of tumors on CT and MRI scans. Maybe it's not about epidemics, but fighting diseases nonetheless.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
This is just one piece of the puzzle. Having a cheaper (less SMS messages) way of running surveys won't by itself thwart the spread of the epidemics. But we can use ML/DL to facilitate development of new drugs, or model possible modifications to the environment (e.g. wiping out mosquitoes, which I mentioned in a previous comment).
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
As I said, CNNs in Medical Imaging - regardless of the specific area, for at least two different tasks - segmentation (feature detection) and classification of images / image patches.
It can be used to develop tests which have many input variables and could be performed for screening without the presence of a doctor (maybe using some portable kit and a medical technician).
A few thoughts on the article - the purpose was not the accuracy, which did not surpass the Supervised Learning (SL) baseline for the full test (8 questions), but limiting the number of queries which lead to a correct classification.
The SL was working with limited data and fixed question, ranked by highest correlation'. Well, what if the top 2 and top 4 features are correlated with each other? The fairer comparison would be to check all combinations of the input variables, and take the best accuracy and then compare it to RL. The SL has "average k" equal to k by definition, which is "max k" for RL. So there is no way that they could have gotten worse number of queries using RL.
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteHello, unfortunately I do have any example in that field. I know an examples where Deep Learning helps to track whales' travel or optimize fishing.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
The biggest limitation that can make that impossible is the complexity of the problem. To analyse effectively environmental disasters you have to take into account billions of variables and process it. Simplified model could not be enough to prevent it efficiently.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
I believe we are able to use Deep Learning as a source, which would be efficient warning system - that can predict disaster using selected (filtered)parameters.
Hi, thanks for the comments. I have to agree with you. The problem is quite complex. It is a good idea to develop some kind of an early warning system - to find certain variables which are leading indicators - forerunners of the epidemic occurrence.
DeleteI tried to google for other examples but I haven't found anything I could share with you.
ReplyDeleteAs for the second question, I think that simulation techniques like multiagent systems are better in the case of epidemy. Knowing some basic features of a pathogen and for example mobility of a given group, it's possible to sufficiently model spreading of the disease to prevent an epidemic.
For me, the best use case for deep learning in this field would be a medical self-examination. People living in regions where the risk of infection is high could take pictures of themselves (access to a camera and other necessary devices is a different story) and then those pictures would be automatically analyzed for early detection of signs of infection.
Hi, thank you for sharing your investigation. I think you are right about using simulation techniques. It will undoubtedly help in measurement of the problem as well as to indicate how fast will the disease proceed.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteUnfortunately I didn't hear about similar examples of use Deep Learning, but it's really interesting and this article show that it is possible to warning and help people in this way. Moreover cost of that help is very low considering that human life is at stake...
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
Preventing is strong word. I think that it could only show us scale of damage. Any algoritm wouldn't get us all warning. All of the pathogenic bacteria constantly mutate and way of their change we couldn't recognise before their will be tested.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
Only like warning systems. This is too unpredictable.
Hi, thanks for the answers. You are right about the new types of bacteria - we are not capable of forecasting the way it mutates and how fast will it attack.
Delete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteThis is not my area of interest so unfortunately I did not recall any other examples.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I hope yes, but as we know we need to have proper and large amount of data which will be used for prediction. And as we know there is hard to gather good quality data.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
I thing deep learning may be used for warning systems.
1. I know it is widely used to predict all the issues connected with hurricanes, storms, whirlwinds and other natural disasters. When I was in Italy, visiting Pompei (the ancient town, discovered two centuries ago, unearthed from under the volcanic dust), the people from there told me they always know about every expected explosion. That helps to prepare and gives possibility to avoid all the danger. Another thing is, if people want to use this knowledge, because - what’s interesting – in Pompei there is a lot of people who never evacuate…
ReplyDelete2. I am optimistic in the respect of the preventing outbreaks’ capability. I am sure some problems will disappear only because of communication’s improvement in the future. The information spreads faster and faster and predicting the natural environmental processes is one of the most important advantages of new technologies nowadays. Paradox is that it helps humans adapt to the world they live in, to survive.
3. The most interesting in this respect is that every new case uses the older ones and that’s why the issue given is less and less dangerous in the future. Another problem is, that new ones appear.
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteI have heard about Deep Learning in medical image analysis. Maybe it’s not an epidemic example but I think that related to it.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
According to the article, mentioned agent is able to predict with 80% accuracy so we can say that it is an effective method. I believe that there is a possibility to improve that agent (I mean its algorithm) to achieve a higher level of accuracy. It seems that Deep Learning is a very good way to recognize epidemiological diseases.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
I don’t have any other ideas because it’s not my cup of tea. However I think that presented possibility of Deep Learning usage is very promising.
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteNo, I do not. This is not my area
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I think yes. We can find patterns that may alert us in proper time
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
It can help us in finding patterns and solutions that will allow finding new drugs
1. Do you recall any other epidemics?
ReplyDeleteThe method is quite interesting but I have not encountered similar solutions.
2. Are we capable of using dry learning out as malaria effectively with Deep Learning? The article shows that yes. :) Generally, deep learning can work in many areas successfully!
3. What are the other possible Deep Learning?
This is not my area of interest so I can not answer.
Unfortunately, I can’t give another example but I think that Deep Learning has enormous potential and can be used in many different areas. The biggest problem from our point of view is the division of factors into these important ones and those connected with information overload. I don't know - maybe I'm wrong but it seems to me that in the future, Deep Learning will not have any problems with it.
ReplyDelete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDelete2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
I am not sure if it was Deep Learning application, but at the last conference I was, two research teams presented their work on an applications following and predicting the tuberculosis cases. They were following internet (e.g. social media) to collect the data about the disease. The main problem as I remember was incompleteness of the data, as there were the areas of the country that Internet was almost not used by the people living there.
In my opinion we are capable of preventing outbreaks of malaria (or another disease), but we have to have complete data for it.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
I am not Deep Learning specialist, so it is difficult for me to answer this question :)
1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteI think that scientists tried to predict earthquakes. I'm not sure if they have succeeded or not.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
As with everything helpful, I believe it can, but it's a long shot in this case.
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
In my honest opinion IoT will be more helpful. Maybe if we combine deep learning with sensors from IoT we will discover a new way of preventing things like you have mentioned. Who knows?
Thank You Katarzyna for your article and practical example of the using Artificial Neural Networks to predict disease outbreak. I am going to remember this example when I am going to talk about ANN. I guess that probably in meteorology similar approach is used but frankly speaking I haven't got any idea about similar appliance in medicine. I recon that IoT or other connected sensors could be another weapon to fight other diseases outbreaks or epidemic.
ReplyDelete1. Do you recall any other examples of application of Deep Learning approaches/research in fighting epidemics and disasters?
ReplyDeleteAs I know deep learning is used for RTG analysis. If there is a common disease this analisis may help to establish proof of existence of epidemic before other methods do so.
2. Are we capable of preventing such outbreaks as malaria effectively with Deep Learning?
80% of accurate predictions seems to be rewarding,
3. What are the other possible Deep Learning approaches which may be helpful in fighting such occurrence?
One of my area of research is analyzing of air transport networks. One of sub-topics is epidemic spreading due to air network. Deep learning may help to predict where the problems will ocure first and to estimate the scale of the phenomenon.