kalyan0510 / activelearningtechniques Goto Github PK

Comparison among Various Active Learning Strategies and CNN on RFI data

Kalyan G - 140001011

Guide - Dr. Aruna Tiwari

Active learning is a special case of semi-supervised machine learning in which a learning algorithm is able to interactively query the oracle to obtain the desired outputs at an unlabelled data point. Since the learner chooses the examples, the number of examples to learn a concept can often be much lower than the number required in normal supervised learning. The way learner chooses the examples is called Query Strategy. Every active learning algorithm needs a query strategy.

Radio-frequency interference (RFI) is an Electromagnetic interference (EMI) when the radiations are in radio frequency spectrum. It is a disturbance generated by an external source that affects an electrical circuit by electromagnetic induction, electrostatic coupling, or conduction. Radio frequency interference (RFI) often occurs as short bursts (< 1ms) across a broad range of frequencies, and can be confused with signals from sources of interest such as pulsars.

We will compare the learning ability of active learning algorithm when implemented over  the following query strategies:

Uncertainty sampling

• Entropy Sampling Query Strategy

• Least Confident Query Strategy

• Random Selection 

Query by committee

• Kullback Leibler Divergence Query Strategy

• Vote Entropy Strategy

And also compare its precision with a convolutional neural network model 

RFI-Detection

Radio-frequency interference (RFI) is an Electromagnetic interference (EMI) when the radiations are in radio frequency spectrum. It is a disturbance generated by an external source that affects an electrical circuit by electromagnetic induction, electrostatic coupling, or conduction. This signal is converted into spectrogram (Image) using Short-time Fourier transform. This image can be used to determine if there exists a disturbance (RFI) or not.

Data without RFI :

Here I have used simulated spectrograms for training our active learning model. Our algorithm have to learn identifying patterns in an image, we can find which query strategy is good at doing this job of identifying patterns by comparing among them.

Results are in the pdf....