Arcane Reveries

This is about Mani, a developer, a curious mind and an amateur astronomer. I heard that you wanted to know me, so here you go.

A Survey of Supervised Learning Algorithms

Feb 10, 2019
Reading time ~20 minutes

Here we analyze the performance of a few supervised learning algorithms in solving classification problems. We don't discuss the inner workings of the algorithms but only their effects and behaviors under different parameters when acting on different datasets.

Ubiquitous Technologies for Healthcare and Wellbeing

Jan 14, 2019
Reading time ~15 minutes

There are two kinds of systems in use in Healthcare with different environmental setting and anticipation but with the same purpose of patient care, point-of-care situations and patient-centric sensing. We will discuss and reflect upon how computing is aiding both in different ways for various different outcomes.

Traveling Sales Person

Dec 12, 2018
Reading time ~22 minutes

We discuss the Traveling Sales Person problem and its complexity. There being no polynomial solution to the original problem in time different techniques have been applied to by using the practicality of domain problems. We explore algorithms that are Approximate but with a guaranteed bound in error and Randomized ones...

Heist Complexity

Sep 13, 2018
Reading time ~8 minutes

We'll see about heist complexity and the ways to compute it efficiently.

Learning to Land on the Moon

Mar 3, 2018
Reading time ~12 minutes

In this experiment we implement a neural network based agent to optimize a future reward function for a sequential decision problem and explore the numerous intricacies involved in the problem. We present an agent that successfully learns to consistently steer towards the goal in a simulation of a module landing...

On Random Walks and TD Learning

Feb 19, 2018
Reading time ~8 minutes

Temporal difference learning methods introduced in 1988 by Richard Sutton are the foundation of Reinforcement Learning algorithms, although the context was different. The celebrated paper presents the idea of TD learning as a general framework to achieve faster learning in prediction for sequential decision-making problems. In this report we’ll reproduce...

Fishy Cyber Attack Detection in Industrial Control Systems

Apr 28, 2017
Reading time ~7 minutes

Cyber attacks have become serious threats to Industrial Control systems as well. It becomes important to develop a serious threat defense system against such vulnerabilities. For such process control systems, safety should also be assured apart from security. As unearthing vulnerabilities and patching them is not a feasible solution, these...

Possibility of Possible Worlds

Apr 21, 2017
Reading time ~12 minutes

We humans have an innate tendency to question everything. We are always drowned in an infinite ocean of provocative thoughts, but a small stream keeps on asking, why is everything the way it is? Why isn’t it the other way around, where the Sun sets in the East? Is it...

Performance of Dynamic Branch Predictors in Processor Pipelines

Apr 20, 2017
Reading time ~9 minutes

Branch Instructions are critical to the effectiveness of a deep pipeline, as these instructions can lead to stalls in the execution flow and result in high penalty costs. The performance of the CPU decreases with increasing pipeline stalls. To better extract parallelism we need to predict, pre-fetch and start executing...

On segmentation of Ultrasound Images of the Neck

Nov 16, 2016
Reading time ~12 minutes

The goal of any image segmentation algorithm is to represent the image in a more meaningful manner and to identify structures and boundaries around them. In its simplest form segmentation is assigning class labels to every pixel in the image. We’ll be segmenting ultrasound images of a region around the...

Inversion of Control

Oct 27, 2016
Reading time ~5 minutes

A sample implementation of the Inversion of Control principle with a comparative analysis of the approach with a traditional formulation of the same problem.

Who's Hat Am I Wearing?

Sep 26, 2016
Reading time ~4 minutes

Nicod's criterion is one of the fundamental concepts of the theory of confirmation. It is also center to the famous 'Raven Paradox', and attempts have been made to debunk it. One such counter example was provided by Rosenkrantz to show that instance confirmation leads to paradoxical states. However I believe...

Cats or Dogs

Apr 11, 2016
Reading time ~4 minutes

In this experiment we'll train a machine to distinguish between cats and dogs. We'll be using Support vector machines for the same.

Visual Word Representation

Mar 8, 2016
Reading time ~2 minutes

Do images speak?

Harris Point Detection and Image Matching

Feb 25, 2016
Reading time ~2 minutes

How well can we compare two images?

Unsupervised Learning of Manifolds

Jan 26, 2016
Reading time ~5 minutes

We will learn the structure of space for a mobile robot, based on observations (i.e. images from the robot's camera) and nothing else. We will be using the isomap algorithm to discover the lower dimensional manifold embedded in the hyper dimensional space of the image pixels generated by the image...

Psychotherapist chatbot - Doctor.el

Jan 11, 2016
Reading time ~4 minutes

We study the behaviour of the Emacs Psychotherapist programme and thereby make some changes to the source of program doctor.el and hope to bring about some change in the way how the program interacts with us.