Page Not Found
Page not found. Your pixels are in another canvas. Read more
Sitemap
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
About me Read more
Page not in menu
This is a page not in th emain menu Read more
Posts
Coming soon…
Published:
Something new is cooking, will be served soon…
- Previous blogs:
- Tips and tricks for Jupyter notebook issues blog
- How to deal with AWS (Amazon Web Server) blog
- Bridge the gap between the remote and local system blog
- Beginners guide for text preprocessing in NLP blog
- Training Word2vec using gensim blog
- BERT Text Classification using Keras blog
- DistilBERT Text classification using Keras blog
- Docker tutorial for Data Science blog
portfolio
Portfolio item number 1
Short description of portfolio item number 1
Read more
Portfolio item number 2
Short description of portfolio item number 2 Read more
publications
Breast Cancer Histopathology Image Classification and Localization using Multiple Instance Learning
Published in IEEE WIECON 2019, 2019
This paper presents a multiple-instance learning based method for classifcation and localization of breast cancer in histopathology images. Read more
talks
Image Basic editor
Published:
- Designed an image processing GUI in MATLAB, with functionalities such as load, save, undo, reset, quit and operations including blur, sharpen, histogram-equalization, gamma correction and log-transform.
Read morePothole detection and avoidance system
Published:
More information here This project is based on recognizing potholes and avoiding road accidents.
- Project description:
- Implemented semantic segmentation network to recognize pothole to avoid road accidents. Performed data preprocessing using image labeler tool in MATLAB for pixel level classification.
- Trained the network with 240 images and extensive comparisons with techniques such as edge detection and thresholding on dataset images. Achieved validation accuracy of 62.5%. The resources for the talk are here.
Design and Architecture of convolutional neural network
Published:
This talk presents an overview of the domain of deep learning.
- Explored the design principles and difficulties in training large scale deep neural networks such as initialisation strategy, batch normalisation, dropout and residual connection.
- Implemented activation functions such as sigmoid and ReLU along with hyper-parameter tuning.
- Evaluated loss functions such as cross entropy, weighted cross entropy, dice, inverse dice and focal loss with corresponding changes in the output.
Credit card fraud detection system.
Published:
- Project description: Developed an SVM model from scratch for classification of fraudulent credit card transactions on a dataset taken from Kaggle.
- Objective : Training Support Vector Machine for detecting Credit card fraud using given data.
- Method Used: 200 samples are randomly take from the data file (100 each for positive and negative class) five times.
- Trained the Model 5 times and obtained respective accuracies using 80% of the data for training and 20% for testing split.
- Achieved accuracy of 92.5% by formulating an algorithm for convex quadratic optimization problem of dual form of SVM.
- cvxopt library has been used for finding out the optimum solution for Lagrange Multipliers.
Automatic Pixel-wise Labeling
Published:
- Project description:
- Developed a softmax classifier for classification of hyperspectral remote sensing images and trained the classifier using maximum likelihood criteria.
- Achieved accuracy of 83.5 % on Indian Pines dataset having 224 channels and 16 ground truth.
Daily stock price prediction of NIFTY index and various campanies listed in NSE.
Published:
- Project description:
- Processed past 10 years financial data from Yahoo Finance to extract technical indicator features such as RSI, MASI, and EMA. This project focuses on predicting stock price trend for a company in near future.
- The prepared dataset looks like the figure shown below .
- Proposed Methods:
- Linear Regression
- Regression using SVM
- ANN Regression
- Created Indictor functions:
- Technical indicators use statistical properties of the present and previous samples, so it can be considered as more related feature as it becomes easy to uptrend and downtrend.
- Technical indicators used:
- Relative Strength Index (RSI)
- Measures speed and change of price movements.
- RSI = 100 – [100 / ( 1 + (Average of Upward Price Change / Average of Downward Price Change))]
- Generally, oscillates between 0 and 100, we consider overbrought above 70 and oversold below 30.
- Money Flow Index (MFI)
- Related to RSI but incorporates volume too where RSI considers prices only.
- Typical Price = (High + Low + Close)/3
- Money Flow (not the Money Flow Index) is calculated by multiplying the period’s Typical Price by the volume.
- Money Flow = Typical Price * Volume
- If today’s Typical Price is greater than yesterday’s typical Price,it is considered Positive Money Flow.
- If today’s price is less, it is considered Negative Money Flow.
- Money Ratio = Positive Money Flow / Negative Money Flow.
- Exponential Moving average (EMI)
- SMA = avg of price data, EMA = more weight to data which is more current.
- EMA is more sensitive to price movement and used to determine trend direction
- EMA = (K x (C - P)) + P
where, C = Current Price
P = Previous periods EMA (A SMA is used for the first periods calculations)
K = Exponential smoothing constant.
- Stochastic Oscillator (SO)
- Shows the location of the close relative to high-low range over a set number of periods.
- The default setting is 14 periods, which can be days,weeks,months or an intraday timeframe.
- K=100[(C-L5close)➗(H5-L5)]
C=the most recent closing price
L5=the low of the five previous trading sessions
H5=the highest price traded during the same five-day pe
- Jupyter Notebook for Technical indicators calculations
- Relative Strength Index (RSI)
- Evaluated extensive comparisons among the metrics MSE, MAE, and percentage error.
- Implemented different models such as linear regression, support vector regression and ANN.
- Support Vector Machine
- Artificial Neural Network
- Support Vector Machine
- Achieved error rate as low as 0.64% using support vector regression in NIFTY index prediction.
teaching
Teaching experience 1
Undergraduate course, University 1, Department, 2014
This is a description of a teaching experience. You can use markdown like any other post.
Heading 1
Heading 2
Heading 3
Read moreTeaching experience 2
Workshop, University 1, Department, 2015
This is a description of a teaching experience. You can use markdown like any other post.