It is one way to learn R and Rpy by reading all the tutorials and documentation. However, the other more indulging approach can be having a go at Kaggle problems.
Prof. Chandy suggested I try to solve two kaggle problems using Python, it's packages and R (or RPy). I got these two really interesting problems that are sure going to test someone's mathematical, hacking as well as comprehension skills.
This blog gives introduction to the two Kaggle problems and start with approach.Though the challenge is very well described at Kaggle, I will give my opinion here.
The two problems given are:
The task is to Create an algorithm to detect North Atlantic right whale calls from audio recordings, prevent collisions with shipping traffic. It is a classification problem, and given 30,000 training set and 54,305 test set, you have to classify each 2 sec audio clip as Whale sound or Non whale sound.
There is a possibility to use Deep learning approaches such as Convolution Networks and Deep Belief Net's here.
Detecting these Dark Matter halos is hard, but possible . If we can accurately estimate the positions of these halos, we can then understand the function they play in the Universe.Given training data, the position of galaxies and their orientation in 2D space, we have to write algorithm to predict one or more than one dark matter halos that cause gravitational lensing of the light coming form Galaxies.
The challenge isn't straight forward as feeding the data into a classifier or a regression model. Careful analysis and well thought strategies have to be
used.
I have formed initial approaches to both challenges. Sadly, both the challenges are long over, hence I can't upload my test set results to get scores. Also, solutions and best approaches are documented on several blogs. However I plan to go by own approach and later understand the better ones.
Prof. Chandy suggested I try to solve two kaggle problems using Python, it's packages and R (or RPy). I got these two really interesting problems that are sure going to test someone's mathematical, hacking as well as comprehension skills.
This blog gives introduction to the two Kaggle problems and start with approach.Though the challenge is very well described at Kaggle, I will give my opinion here.
The two problems given are:
The task is to Create an algorithm to detect North Atlantic right whale calls from audio recordings, prevent collisions with shipping traffic. It is a classification problem, and given 30,000 training set and 54,305 test set, you have to classify each 2 sec audio clip as Whale sound or Non whale sound.
 |
| A spectrogram showing whale call of one of the Audio clips. Source: Kaggle description of challange |
There is a possibility to use Deep learning approaches such as Convolution Networks and Deep Belief Net's here.
Detecting these Dark Matter halos is hard, but possible . If we can accurately estimate the positions of these halos, we can then understand the function they play in the Universe.Given training data, the position of galaxies and their orientation in 2D space, we have to write algorithm to predict one or more than one dark matter halos that cause gravitational lensing of the light coming form Galaxies.
 |
| Source: Kaggle description of the challenge. |
used.
I have formed initial approaches to both challenges. Sadly, both the challenges are long over, hence I can't upload my test set results to get scores. Also, solutions and best approaches are documented on several blogs. However I plan to go by own approach and later understand the better ones.
No comments:
Post a Comment