Education
KTH Royal Institute of Technology 2016 - 2018
Stockholm, Sweden
- M.S. in Machine Learning
- Thesis supervised by Dr. Arvind Kumar
PES Institute of Technology 2010 - 2014
Bangalore, India
- B.E. in Telecommunication Engineering
- Thesis supervised by Dr. Koshy George
Publications
A Multiple Model Approach to Time-Series Prediction Using an Online Sequential Learning Algorithm
Koshy George, Prabhanjan Mutalik.
IEEE Transactions on Systems, Man, and Cybernetics: Systems.
[PDF]
[BibTeX]
[www]
Online time series prediction with meta-cognition
Koshy George, Prabhanjan Mutalik.
International Joint Conference on Neural Networks.
[PDF]
[BibTeX]
[www]
[repo]
An online multiple-model approach to univariate time-series prediction
Koshy George, Sachin Prabhu, Prabhanjan Mutalik.
Springer, Cham.
[PDF]
[BibTeX]
[www]
Hippocampus as an Echo State Network
Prabhanjan Mutalik
Masters Thesis.
[PDF]
[BibTeX]
[www]
Experience
KTH Royal Institute of Technology 2020 - Present
Research Scientist
PES University 2018 - 2020
Research Associate
Infosys 2014 - 2015
Systems Engineer
Current Projects
Malaria
In collaboration with Karolinska Institute and National Institute of Malaria Research
Connectomics
In collaboration with KTH Royal Institute of Technology
Metacognition
In collaboration with Center for Intelligent Systems, PES University, Bangalore
Previous Projects
- Temporal Sequences in the Brain: An investigation into the role of sequential activity in the brain and its effect on behaviour. The project sought to review the existing literature and identify neural sequences using the Tempotron learning rule.
- Biologically Plausible Learning Algorithms: The project surveys and tests the latest biologically plausible models in deep learning with the aim of finding the optimal model that satisfies the computational requirements simultaneously maintaining fidelity to the biological structure.
- Neuroevolution Algorithm for Time Series Prediction: The aim of the project was to compare Back-propagation Algorithm and Evolutionary Algorithms in the context of time series prediction.
- Translation Optimization based on Language Similarity: Machine Translation (a Seq2Seq TensorFlow model) was optimized based on the distance between the languages on the language tree.
- Symptom Sorter: I have built a symptoms recommendation systems featuring simple probabilistic learning.
- Metacognitive System for Time Series Prediction: The online prediction model consists of a cognitive element (a single layer feedforward network) and a metacognitive layer that controls the hyperparameters of the network.
Programming Languages
- Python
- MATLAB
- HTML and CSS
- Java
- SQL