I am interested in solving pressing problems using interesting techniques and like to implement ideas and built products around them...
I am keen to learn and research in the areas of Machine Learning, Cognitive Computing and Computational Social Sciences!
In this project, we have captured human expressions through questions and responses in pictorial form. The personality is mapped to responses. This work is further developed to associate relationships between clusters and depression. We have proposed a graphical association method for clustering candidates and mapping them to cultures. In graphical association method (GAM) each candidate is represented as a graph with core node and multiple nodes associated with the core node. Core node is the node associated with maximum number of nodes and it changes as more information about candidate is made available. Reinforcement Machine learning is used in this case. Thus, multiple graphs are associated based on closeness among graphs. This closeness is calculated using distance between corresponding nodes. The clustered graphs are represented using a representative graph and mapped to culture cluster. This mapping is used for task selection and recommendation. In this particular application dynamic scenarios are handled using special event related data.
Mentor/Supervisor: Dr. Bradly Alicea
Embedding emotions into product is very interesting and equally challenging concept. Emotions can be embedded in the form of color, text inscription, sound and even pictures. The first and most important part of this project is to determine emotional traits of individuals. This is determined using analysis of expressions, responses and social media posts. Based on emotional traits the candidates are clustered and a set of emotions are determined. In some cases, those are context specific. Thus, emotional traits and context is used to identify the emotional requirements. For any product based on emotional traits basic theme is determined while context helps us to determine sub theme. Here Affinity is determined using mathematical models based on closeness factor. Closeness between two or more emotions is determined using combination of Manhattan and Euclidean distance. In static model basic emotional trait based features are used. The context-based features can be dynamically included in the product. The work is in the area of Computational Psychology and Computational Behavior. The association between products and emotions is used for recommending products and context specific emotions are embedded in products.
Thinking and thought process measurement is very subjective. Can we find out and rank thought process of students, faculties or researchers? In case we could do this it can revolutionize the education system. It can change the way we measure and rank performances. Measuring and Ranking thought process remains a challenge. We teach Physics, we teach Math but we always fail to teach thinking. There are psychometric tests to evaluate thinking. But these tests are very subjective. Thoughts have associations with actions. Capturing this thought process is what we are trying to achieve using this innovation.
Mentor/Supervisor: Dr. N. Rajopadhye
The path followed by consumer is depictive of his behavioral pattern, his likings and needs. Overall path tells us about his buying pattern and can help us to learn about articles he/she is planning to buy. Associations among these traversal paths could help us to predict traversal path of similar customers. This can allow us to suggest personalized paths for customers and also to rearrange articles dynamically based on majority of customers expected on a particular day. This is based on algorithm for associating traversal paths and consumer psychology to predict behavior of customers. The traversal path is represented as a graph. The major nodes are the products of significance where buyer spend time more than 1 minute. This traversal path decides association among objects mapped to the buying patterns of individuals. In this project we clustered a number of such traversal paths and associated it with customer representative data. This information is used to determine traversal path of the same customer or similar customer during their visits.
It is the context of task, behaviors of individuals and above all constitution of the team in that scenario contribute to the outcome. The proposed technique is based on convergence of multiple context vectors representing computational behaviors. Learning based on these vectors helps us to select the optimal combination. The Context Vector Convergence (CVC) of Behavioral Vectors helps in deriving the actual effect of two vectors in overall team performance. The personality vector is used to derive behavioral context while mission vector is used to derive the scenario context. Thus, team is selected so that negative emotional impact on team members is minimized. The algorithm of context vector convergence is proposed in this work.
Mentor/Supervisor: Prof. M. Marathe
News unknowingly creates desired or undesired psychological impact. Right from title, presentation and sequencing to filtering and personalization - machine learning, and cognitive sciences can play a key role in news computing and processing. This project looks into research carried out in the area of cultural and news computing in elaborate way and proposes a model for personalized sequencing and presentation of news. This personalized newspaper aims to present newspaper of your liking and suitable to your emotional makeup so that the overall desired impact will be achieved. Further it helps in countering depression that may result due to negative news. In this project personal data of reader captured through our technique of responses to pictures is used to determine his/her cultural and emotional traits. The news articles are ranked with reference to user and even collated to derive desired impact. So far we have worked on expression analysis and news mapping.
Mentor/Supervisor: Prof. M. Marathe
Back end to classify patients based on properties and mapping the patients to sequence of events. Special help module for emergency. - Android, Java, python Features: Alerts, Reminders based on situation, Context based event trigger
Mentor/Supervisor: Prof. A. Bhadgale
Problem: Transfer of Employees
Solution: Transfer based on skills, suitability and context. Automated transfer module.
Learning based on human inputs,
Algorithms: Statistical machine learning and parametric association
Mentor/Supervisor: Prof. M. Marathe
Based on the concept of Vichardhara, a provisional patent is filed in Aug 2017.
Filed an Indian patent on July 2018 E-2/1489/2018/MUM and PCT in July 2018.
Hrishikesh Kulkarni, "Contextual Data Representation Using Prime Number Route Mapping Method and Ontology" IEEE Conference, ICCUBEA, June 2017
Worked on Context based personalized analysis of individuals to decide premiums and claims. Further worked on developing an intelligent algorithm and culture based analysis so that the agriculture and health care insurance benefits could reach to Bottom of Pyramid. The core part of this project is to mine farmers' intent. In intent mining farmer association, farmer context and expression are used. We have used periodic responses from farmers along with contextual data to mine their intents. Thus, farmers are mapped to the most suitable schemes.
Mentor/Supervisor: Prof. (Dr.) S. D. Page
The LTO object file is a regular elf file with sections containing LTO byte-code. A LTO object file contains various sections for storing command line options, symbol table, global declarations and types, function bodies in GIMPLE, ipa pass summaries, ipa references, static variable initializers and the call graph. There are couple of limitations of the byte code format: 1] It is not self descriptive, which makes it harder to debug. 2] The byte code is essentially a "serialized" version of in-memory representations, which makes it prone to break across versions.
The purpose of this project is to create a dump tool for easily analyzing LTO object files similar to readelf or objdump -d for regular ELF object files. Link Code
During this internship I got opportunity to play with data. During this internship I used different existing clustering techniques to cluster behaviors. We used different statistical methods. I could take part in proposing a new algorithm for intent mining. In this internship data of different teams and performance of individuals and teams is used. The data is analyzed considering different attributes. I have used different keyword and key-phrased based associations. At later stage we build a node association map. This map could yield better results than traditional approaches. Apart from this core work I also got opportunity to work on different aspects thought process mapping.
Hosting duty is working on developing and hosting websites along with mining website data. During this internship I got opportunity to make my hand dirty with actual website development work and basic web mining activities. For web mining I used all off the shelf algorithms like TFIDF. I used cosine similarity to associate data. The aim of this work was to detect exceptions and anomalies in information. We used signature-based approach for the same.
Mentor/Supervisor: Nenadd Chandorkar
Striving to solve social problems using Research and Technology. If you would like to discuss on Cognitive Computing over a cup of coffee, feel free to drop me a mail at firstname.lastname@example.org