Machine Learning Intern At Sabudh Foundation

Tried pretty hard for the collaborative model to make the dataframe in memory regarding the user profiles and got a new idea after whole day of trying which seems even more difficult to apply than in theory. Idea is to create a dataframe with user ids and the user profiles and each time read the user id and if the user id already exists then apply some function to overrite the data with the averages of the both and in  case it doesn't already exists create a new user id and its profile. Gotta try it tomorrow. Good Luck to me. :"D

As i completed the generative model first in the whole batch, i am now looking forward to complete the Collaborative Model that generates the Trending and nearby articles and will be working on it tomorrow. That would mean ending of the News Recommender System Model and will be working on something new, maybe and hopefully Post Tomorrow. Be Posted.

Worked on Generative model as we got another assignment for the day which is gathering the NN news which also is trending to the most user profiles and showing them which works as a collective news generation model.

Completed the Bot2 for News Recommender System today and got the generative data model work today.  Will complete that work by tomorrow. 

I worked on Bot 2 for the news recommender system today. Learned various new topics : -Pickling/Depickling -Serialization/Deserialization Had a issue with cosine similarity with dimensions not matching and giving an error but was solved finally by mapping the indexes of the array to list twice.

Did image cropping today from the video and the car is getting cropped in project. Completed BOT 1 of news Recommender project which now does things randomly and according to the number of news the user wants.  Also helped my fellow friends in completed their BOTs.

LDA completed.. BOT 1 working and providing with random news articles based on KMeans clustering.. Worked on cropping an image which was quite a bit of a mess with assertion errors and stuff.

Worked on the news recommender system Bot one. Worked on LDA.

Started ISB videos from the beginning to revisit the concepts taught and to get a better understanding of the concepts skipped or the concepts those were not fully understood. Watched 1 full video which is quite the length of a movie and which has a lot of stuff. Transferred the project work from live video working to input video work which will help in the POC to be provided to the client.

Worked on News Recommender project Bot 1 and worked on stemmer and lemmatizer and different kinds of stemmers. After getting poor results with stemmers removed stemmers altogether and used lemmatizers instead. Then did clustering with Kmeans and used an elbow curve method to find the number of clusters and used KMeans clustering with 5 clusters and found the clusters that means i got my news divided into 5 different clusters which i could recommend to user by my Bot 1. Worked on project and completed the task that was to be completed today that is to detect the image of the customised when it passes a line or say a point i want it to pass which will help me in the trigger of my algorithm.

Did further videos of Spatial Data Science and the project of news recommender using NLP and tfidf transformer and worked on it further. Worked on the project further on custom detection of cars on my live video stream and finally was able to get the detected picture with the customized detection. Tomorrow will work on how to take a picture with motion and after detection from a certain point.

Worked the spatial data science course of JOON HEO who speaks a funny english. Did 3 weeks lectures in a single day and then studied upon Word Embeddings and word2vec Google API for the same. Then worked on project applying a fillypoly or a line to a live video or taking out frames of a live video and working upon them.

As we proceeded with the unsupervised machine learning and natural language processing we began with the spatial data science and watched Sarab sir's lectures of bayesian learning and clustering. Meanwhile we got a project "News Recommend-er System " which recommends the news to the users automatically as per user's interests and using the NLP and 2 bots which carry the first time users and existing user details and recommends the news.