DeepFragrance a peek into the olfactory chemical space

Fragrant molecules are omnipresent in our daily lives, from food over lifestyle to hygiene. Yet, their design is an iterative and time consuming process. Recently, machine learning models were developed capable of predicting the perceptual qualities of 476 molecules using its structural properties (Keller et al.(2017)). It demonstrated the feasibility of the approach, but covered only a microscopic fraction of the total “fragrance-space” consisting of about 60M molecules (Ruddigkeit et al.(2014) ).

We leverage recent advances in deep learning to expand the search space of potential odorants, and to identify new candidate fragrant molecules. As a first step, we have compiled a dataset consisting of ~109 small molecules, including ~107 fragrance-like molecules with unknown odour, and an in-house collection of molecules related to human odor space (Kumar et al.(2015)).

We apply deep learning by training a deep autoencoder, a form of unsupervised learning, to form a representation of fragrance space. We initially trained word2vec and doc2vec models on ~52k molecules. The features were obtained by NLP strategies, and molecular fragments obtained from cleaving of the SMILES notation along synthetically accessible bonds.

Preliminary results indicate that fragrant molecules occupy a certain extent in chemical space, indicating a region in which potential candidates of previously unknown fragrant molecules can be found, effectively limiting the search space. In future work we aim at scaling up our machine learning efforts to the full size of the available data, and combine our unsupervised approach with supervised methods to predict the scent of new fragrant molecules.

This research is funded by the Royal Society, UK as part of Newton International Fellowship grant NIF\R1\181328.

References

Keller, Andreas, et al. Science 355.6327 (2017): 820-826.

Ruddigkeit, Lars, et al.. Journal of cheminformatics 6.1 (2014): 27.

Kumar, Ritesh, et al. PloS ONE 10.10 (2015): e0141263.

A mobile application for effective relief works during floods

Overview

Mahatma Gandhi once said “I would say that if the village perishes India will perish too. India will be no more India. Her own mission in the world will get lost“. We live in interesting times where the realities of people and the establishments have become multi-layered. While on one side, immense infrastructural and technological developments have taken the roots in massive urbanisations, on the other side, we see some of the villages devoid of even the basic facilities. This has led to a widespread migration of people in the search of better jobs and livelihood. It has created more problems than mitigating it, creating another class of people called as urban poor. It has also jeopardized the village economy by affecting  the number of workforce, environment, and in a market driven economy, lessening the chances of private investment. The loss of investment is a vicious cycle, the lesser the money invested, the lesser chances of employment, the more people migrate to urban settlement and the cycle continues. The government apathy has further made the situation worse. While there are large and very interesting schemes to kickstart the investment cycle in the rural sector, one does not have to go too far to see how they are faring. The primary reason for such a spectacular failure till now, can be directly attributed to the bad or even the worst conditions of basic public infrastructure like roads, schools, dams, health centers and myopic solutions to major structural and environmental problems.

The Indian sub-continent is fortunate to have three to four different seasons and a huge reservoir of water in the form of perennial and seasonal rivers. The perennial rivers are bolstered by the reserves of glaciers in Himalayas and water sinks in the south. The seasonal rivers are filled in the rainy seasons. With many tributaries and distributaries of the major rivers, the whole subcontinent except some of the parts gets covered by these rivers. The whole thing said, we still face major water shortages in the peak summers throwing rural economy in jeopardy and has also been associated with large number of farmer suicides. The reason is very simple to state, due to huge manual interventions in the form of deforestation and burden of ever burgeoning population on water reservoirs, there  is a non-equitable distribution of water, which leads to scarcity.

Flood is one of the most common occurring phenomenon plaguing our country. Almost 75% of the rainfall is concentrated in the four months (June-September) of monsoon resulting in the rivers carrying huge amount of discharge in this season. The prime reason for flooding is inadequate drainage systems in residential areas,  unchecked civil development and most importantly no provision for distribution of excess water from the overflowing waters in the rivers, creeks and other such sources. It is also compounded by the problems of sediment deposition and sea tides in the same season. The himalayan rivers tend to carry huge amount of silt in turn causing erosion of the banks in the upper regions and deposition in the lower areas. The Brahmaputra and the Gangetic plane are the most flood prone areas. Besides these, west flowing rivers like Narmada and Tapti wreak havoc in the Central India and the Deccan along with the  the east flowing rivers like Mahanadi, Krishna and Cavery. The area liable to floods is more than 40 million hectares. The annual average cropped area affected is approximately 8 million hectares.1,2 The last ten years (period from 1996 to 2005 ) saw an average annual flood damage of Rs. 4745 crore as compared to Rs. 1805 crore, the corresponding average for the previous 53 years.2 On an average 1600 lives are lost and crores of peoples are affected by the flood. The measures to mitigate the miseries of people in such situations need a very strong and broad political, social and technological spectrum to be addressed.2 Figure 1 shows the flood zone of India.

floodzone

Figure 1 : Flood Zones of India (http://www.ndma.gov.in/en/zone-maps.html )

Flood relief and disaster management is not in any of the three lists formulated by the constitution makers probably because, it involves multiple subjects already dealt within  the lists. Since it does not figure in the any of the lists exclusively, the parliament has the competence to legislate on this subject but conventionally because of the nature of the matter, the state governments  generally handle the responsibility. Since the entry 23 in the concurrent list contained ‘Social security and social insurance, employment and unemployment’ , the parliament enacted the Disaster Management Act, 2005 (although all aspects of disaster management can not be said to be fully covered by this entry). It provides the National Disaster management Authority (NDMA) the role of planning, coordinating, monitoring and providing assistance during a disaster. It also provides the executive functions relating to implementing and disaster response. The actual implementation should lie with the state governments.

The most important aspect of any disaster management planning and relief works is information dissemination and assessment of the damage along with the information to the affected areas. The NDMA  issues do’s and dont’s in case flood is likely to hit an area.3 Some of them are given below.

  • Listen to the radio or television for information.
  • Be aware that flash flooding can occur. If there is any possibility of a flash flood, move immediately to higher ground. Do not wait for instructions to move.
  • Be aware of streams, drainage channels, canyons, and other areas known to flood suddenly. Flash floods can occur in these areas with or without such typical warnings as rain clouds or heavy rain.

Problem Area

The state of Bihar is landlocked by Nepal in the north, Jharkhand towards the south,  West Bengal in the east and Uttar Pradesh in the west. There are several rivers which criss-cross it, in turn making it one of the most fertile planes in the world. To name a few, Ganga, Kosi, Punpun, Falgu, Karmanasa, Durgavati, Gandak, Ghaghra etc. Almost 85% of land of the state is under cultivation. It also receives heavy rainfall during the month of June to October. 4                 

The relationship of flood and Bihar has been an eternal one. Almost half of India’s average loss is accounted for by it. The most dreadful memory of flood has been in 1914, when both Bihar and Bengal faced it. This was followed by massive earthquake in 1934 followed by flooding. Since then floods have been a frequent occurrence but its frequency has increased in recent times. Year 1979 again saw massive damages when lakhs of people were displaced and several thousand lost their lives besides loss to livestocks and property. The state has faced infrastructural losses worth crores of rupees.

Recently, in 2008 flood fury gain hit Bihar submerging half of the land mass. More than 30 lakh people spreading over 1500 villages in 16 out of 37 districts were affected.  Araria, Saharsa, Supaul and Madhepura were the worst affected districts. This was followed by two successive droughts but the floods came back again in 2011 when almost 100 villages were inundated by Bagmati river.5,6  Most of the damages in Bihar has been caused by the Kosi river, which is also called as the ‘sorrow of Bihar’.  It is a major tributary of the river Ganges. It drains about 60,000 km2 of eastern Nepal and southern Tibet before entering Bihar. It is known as the Sapt Kosi or “Seven Rivers” in reference to its seven tributaries: Indrawati, Sunkosi, Tambakosi, Lihku Khola, Dudhkosi, Arun and Tamur in the north of India-Nepal border. The three main tributaries Arun, Tamurand Sunkosi join the river at Tribeni. Towards the downstream of Tribeni, the Sapt Kosi flows through a very narrow area of 11 Km, before spreading into the Gangetic plane. This causes a sudden decrease in slope and formation of inland delta.  Interestingly, the river has shifted more than 100 km westward in the past 200 years.  

The primary reason of incessant floods is the increased conversion of forests to agricultural and pasture land in the middle hills of Nepal. The river Sapt Kosi has seen an increase in the annual run off from 1950-1980 along with an increase in rainfall. The cause of damages can also be attributed to the increase of population in the floodplains of the river.  The state government has built about 3000 kms of embankments, but the flow of the river has grown 2.5 times resulting in the failure of embankments in every flood.

Geographical Location

Begusarai is centrally located in North Bihar. At first in 1870, it was established as a sub-division of the Munger district. Later in 1972 it was named as a separate district.  The history of the  name of the district is contentious one, some say it was the Begum Bhagalpur who used  to visit “Simaria Ghat” (holy place on the banks of the Ganges) for a month of pilgrimage (“Sarai”, a place for taking rest), which later took to the colloquial of Begusarai.  

Geographical features:

Begusarai lies between latitudes 25°15′ and 25° 45′ north and longitudes 85°45′ and 86°36″ east. The town is bifurcated by National highway 31 and eastern railway line going towards Assam. It is surrounded by Samastipur in the north, the river Ganges and Lakhisarai district in the south, Khagaria and Munger in the east and Patna district in the west.  

Many rivers, rivulets, dhars, nalas  drain the Begusarai district besides, chaurs which traditionally act as sinks to preserve rainwater. Some of the rivers are perennial e.g. Ganga , Burhi Gandak , Kosi ,Kareha and Bagmati  whereas, some of them are seasonal viz. Channa, Bainti, Kachna, Monrya and Malti. Almost, all types of streams i.e. Mountain Fed (Ganga), Foothill Fed (Bagmati) , Plain fed(Burhi Gandak) and Mixed fed drains the district. The drainage pattern of the rivers of this region is characterized as Dendritic Drainage Pattern. The Ganga Rivers shows Yazoo Pattern whereas Burhi Gandak , Kareha , Bagmati , Balan , Baintia etc show very high Sinuosity and are typically meandering rivers. Ganga enters Begusarai from Southwest into Chamtha block of this district. River Balan enters the district in the Bachhwara block, Bagmati enters from Bakhri and Burhi Gandak enters from Parihara..

Vulnerabilities

The Begusarai district is in the middle of mid Gangetic plain and most of the area has low lying terrain with slope from the South to SouthEasterly. Geographically, the district can be divided into three flood plains

  1. Ganga flood plain
  2. Burhi Gandak flood plain
  3. Kareha-Bagmati flood plain

Among these the Burhi Gandak and Kareha-Bagmati flood plains are very low lying areas and are very prone to flood. The streams in these regions shift and leave scars of previous channels causing the presence of wetlands ,oxbow lakes and swamps. The entire district of Begusarai falls in the T0 surface also known as Active flood plains.

Ramdiri Village

Ramdiri is a modestly large village located in Matihani Tehsil of Begusarai district, Bihar with total 1379 families residing. According to the census 0f 2011, the Ramdiri village has population of 6712 of which 3605 are males and 3107 are females. The population of children with age 0-6 is 1295 which makes up 19.29 % of total population of village. The average Sex Ratio of the village is 862 which is significantly lower than Bihar state average of 918. The child Sex Ratio is 919, which is again lower than Bihar average of 935.  As per census 2011, literacy rate of Ramdiri village was 56.71 % compared to 61.80 % of Bihar. The male literacy stands at 66.86 % while female literacy rate was 44.75 %. As per constitution of India and Panchyati Raaj Act, the village is administrated by Sarpanch (Head of Village) who is the elected representative of village. Following table shows the aggregate census data of the village. Figure 2 shows the Google map image of Ramdiri village. Figure 3 shows one of the dhabs (embankments around Ramdiri village)

 

Particulars

Total

Male

Female

Total No. of Houses

1,379

Population

6,712

3,605

3,107

Child (0-6)

1,295

675

620

Schedule Caste

94

58

36

Schedule Tribe

0

0

0

Literacy

56.71 %

66.86 %

44.75 %

Total Workers

2,456

1,683

773

Main Worker

1,256

0

0

Marginal Worker

1,200

734

466

 Table 1: The census data 2011, Ramdiri village (source: Census data 2011)

image03

Figure 2: Google Map of Ramdiri Village (source: maps.google.com)

image02

Figure 3 : Ramdiri, Ramnagar Dhab

Problems during floods

Until very recently, flood situation has been very grim in the district. The Village lies on the banks of the river Ganges and year after year is ravaged by the fury of flood. The river in fact rakes up the entire Matihani tehsil in the heavy rainy season. Although, the government has strengthened the river embankments, due to large volume of water and width of the river during heavy rainy season there is a heavy discharge causing destruction of lives and property. There are some villages which get and have been completely destroyed. Besides, death and destruction of human lives and livestocks there is a lot of destruction in the form of material and property. In 2016 only, till now there have been more than 150 reported deaths and a loss of crores of property. While the problems of flood and the associated risks are well known, there is a great delay in relief reaching to people even after the district administration and people get the knowledge of impending flood. This leaves the general public to fend for themselves. There are many well documented problems before, during and after floods like

  1. Loss of lives and livestocks,
  2. Loss of material property,
  3. Trauma,
  4. Danger of spread of communicable disease.

The single most important problem is late arrival of relief works. While the administrative negligence can not be neglected but, there has to be some technological solutions at least before flood water approaches the low lying areas or when the device and communication systems still work before a complete black out.

Solutions

Desperate times need desperate measures. Flood situations when already out need desperate measures when the structural measures have failed. Although, the best solution is a comprehensive outlook to such frequent occurrences, but when the flood has already occurred or there is a clear indication that there is an impending one, there has to be an objective protocol to deal with such situations. Following are some measure which could be undertaken.

  • An Early Disaster Warning System is to be established.
  • A plan has to be prepared to generate awareness among the masses about the ways and means of mitigating the risk of disaster.
  • Technological Solutions:
  • Mobile and communication system can be leveraged as everyone today possesses smart phones.
  • People help people- The data gathering and relief works can be undertook in an objective way using the phone apps.

A number of structural measures can be taken up in the village:

  • Detention Basins: The village area has a number of depressions locally called chaurs which act as detention basins. These chaurs absorb a considerable amount of water of the first flood of the season. No man made detention basins or improvements in natural chaurs has been done.
  • Embankments: All the rivers have been embanked in the state. River Kosi is embanked on both the sides. But there are few gaps in these embankments which reduce its effectiveness. The maintenance and repair of these embankments must be taken into account.
  • Channel Improvement: The river channels can be cleaned and sometimes diverted to other non-occupied or merged with some other streams where equitable distribution of water can be done.

Crowdsourcing the data

Technology can help in alleviating many problems of the society. Here, I propose to develop a mobile app to gather flood data of the specific areas of the village and crowdsource the livestock and trapped human data, so that at least till the communication system works, the lives can be saved. A user will have to enter the details of how many people or livestocks are  there with him/her. The system will automatically locate the source from GPS location and relay it to the authorities as well as provide a real time map to other persons, so that help can be managed locally too.  The system will also provide all the normal do’s and dont’s during floods along with the phone numbers of the corresponding departments of the district.  

Limitations

The application will have the same limitations as any other device. The flood and other such disaster see power cuts and communication failures for a very long period of time. This is a big hinderance for any such device or application.

Conclusion

Technology can be used to solve many of the problems plaguing our villages. We can solve or at least lessen the problems by deftly using the technology.

Pictorial Representation of APP

crowdsourcing

untitled-drawing

 

References

  1. Yadav, R. J. Disaster Management in India: Acts, Policies, Guidelines. Jaipur: Paradise, 2011. Print.
  2. “Floods – National Disaster Management Authority.” Floods – National Disaster Management Authority. Web. 21 Sept. 2016.
  3. “Do’s and Dont’s – National Disaster Management Authority.” Do’s and Dont’s – National Disaster Management Authority. Web. 21 Sept. 2016.
  4. Russell, Jesse. Geography of Bihar. P2013.
  5. BIHAR STATE DISASTER MANAGEMENT AUTHORITY.” BIHAR STATE DISASTER MANAGEMENT AUTHORITY. Web. 21 Sept. 2016.
  6. “FMIS.” FMIS. Web. 21 Sept. 2016.

 

Smell reporting : Do we need a better representative language (1)

 James McHugh in his book Sandalwood and Carrion: Smell in Indian Religion and Culture writes

“We might begin by considering the notion that smell is primitive, bestial and sensual; an animal sense that has been diminished in adopting the human upright posture, and a sense that, if overdeveloped, marks one as a savage.”

Although slightly incorrect, his statements are indicative of our so little understanding of sense of smell. The odor molecules pervade our living universe, literally swept by the air that we breathe in to the sensory organ. The sampling process to de-correlation of the sensory signals, the whole olfactory pathway is filled with mysteries. The first of them being, which molecules interact or which part of molecules is recognized by which receptor ? The others being, Do the similar receptors converge at a single place or is it diverse? Where do these signals go and what do they evoke ? Alas, the only non-invasive and simplest way of knowing what it felt like to smell something are verbal descriptions by subjects or perfumists. These are considered to be notoriously varied in nature to the extent that some researchers have concluded the language abilities to be uncorrelated with olfactory experience. The research and the reports thereby are concentrated in English as vocabulary expression of smell.

The availability of data and modern statistical techniques coupled with computing resources has led to unprecedented advancement in some fields like natural language processing.  In this series of blogs I will be first reporting the olfactory data gathering techniques, the data bases, their analyses and statistical methods to analyze the bigger problem of olfactory stimulus-percepts. I will be attempting to find sense in verbal descriptions of smell using some of the famous techniques and design a co-occurrence network, where I will define the concept of perceptual classes and communities.

I also attempt to gather as many odoriferous molecules as possible, along with their percepts and physico-chemical properties and hence find statistical regularities in perceptual and physico-chemical space. Kindly report any mistake that you may think has crept into the database.

http://odornetwork.com/

The smell graph

Sense of Smell has perplexed many a researchers and artists likewise. Unlike other faculties of human senses it has not been well understood. The Ancient civilizations of Mesopotamia, Phoenicia, Arabia, Egypt, India, Greece and Rome, in some or other ways engineered smell, which we today call it as incense sticks or simply incense. The Indians used it for medicinal and religious practise. The art of incense making was well studied and practised in India and its evidence is in the ancient Vedic texts. There have been some seminal studies relating the different aspects of odor spaces in recent times. I attempt to gather as many odoriferous molecules as possible, along with their percepts and physico-chemical properties and hence find statistical regularities in perceptual and physico-chemical space. Kindly report any mistake that you may think has crept into the database.

http://odornetwork.com/

LokSabha2014: Sentiment Trends

Image

 

The tweets were collected over a period of 4 days (13th, 14th, 15th and 16th of February 2014), total tweets amounting to > 100,000 tweets (using the Twitter APIs). The retweets (tweets which are published once again by people), since in the present study these do not convey any information were removed. This left almost 50,000 tweets varying from 20,000 in a day to 10,000. To maintain or curb the bias of hashtags, only the neutral hashtags and key words were used as filter for these tweets. The tweets collected were thereafter analysed using a sentiment analysis algorithm, which classifies them into positive, negative and neutral tweets. More on the algorithms and hashtags later..

LokSabha2014: Happy Valentine Day AAP

Image

Top negative tweets

  1. Stupid AAP and feku kejriwal are the worst… Cant bear them. Plz resign as soon as possible. #AnarchistCongBJP
  2. “We don’t like corrupt Cong & BJP . But we also don’t like stupid, dumb, impatient, ever complaining never solving, obsfuscative #AAP”
  3. #AnarchistCongBJP @ArvindKejriwal is so useless that he never cared about these poor… Aam Aadmi party my foot! http://t.co/LaSThz1R93
  4. But overall tone of #AAP supprtrs (I was one too) nowadays is “”if u r not with us then u r stupid”””

Top neutral tweets

  1. Will Arvind Kejriwal resign? Defiant AAP says it will table Jan Lokpal Bill despite Lieutenant Governor’s warning http://t.co/D6B0fw7wR7
  2. BREAKING: AAP leader Ashok Agarwal says government should follow constitutional laws while introducing the Jan Lokpal Bill in Delhi Assembly
  3. BREAKING: AAP to launch Aam Aadmi Cricket League for all unsold players http://t.co/hVW1SmGBnf #IPLAuction
  4. “LIVE: Arvind Kejriwal doesn’t want Jan Lokpal Bill to be passed, says Congress leader http://t.co/KYfnghkhLV”

Top positive tweets

  1. “Happy #ValentinesDay to AAP and Congress , the most lovely couple recently married in Delhi ! #Parliament”
  2. Lovely saying listen to me learn from me to #AAP in #Delhi assembly has a touch of reality however funny it sounds @DrGarekar @musicaltrees
  3. “Best couple of this valentine day “”CON + BJP”” hope ur love wl continue till 2014 Parliament Election…..#AAP”
  4. @gdhanu @AamAadmiParty The truth is some excellent work is done by AAP on administrative grounds BUY Bharti’s act is a shame!

 

Disclaimer: An analysis is as good as the data and twitter obviously doesn’t represent the whole country. The tweets above are not for redistribution and do not represent any political affiliation.

Analysis done for purely educational and research purpose.

LokSabha2014: What Twittter sentiment says

Myriad views, strong emotions…the ever expressive Indians have something to say about everything and if, the topics are related to elections, you can not ever stop them from their constant jibber-jabber.  Elections in India bring with themselves a high voltage drama of allegations, counter allegations, some cooked and some made up stories.  The twitter in seconds gets filled with people talking about hundreds of things. We (Ritesh Kumar and Rishemjit Kaur) set out to ask can we use these twisted and some times horrible looking tweets to reflect the sentiments of people towards some of the big political parties (BJP, Congress) and some newbies (AAP).Image

some sneak peek

Stay tuned… we will be putting up daily sentiments in this space..

Disclaimer: An analysis is as good as the data and twitter obviously doesn’t represent the whole country.

Design a site like this with WordPress.com
Get started