Learn More
OBJECTIVE Support Vector Machines (SVMs) are state-of-the-art classifiers in machine learning. However, the prediction cost of non-linear SVMs is very high. The objective of this project is to improve the prediction time of non-linear SVMs. MACHINE LEARNING What is machine learning? It is the ability of a computer to detect and extrapolate patterns(More)
We trained a convolutional neural network (CNN) to map raw pixels from a single front-facing camera directly to steering commands. This end-to-end approach proved surprisingly powerful. With minimum training data from humans the system learns to drive in traffic on local roads with or without lane markings and on highways. It also operates in areas with(More)
Lifted inference algorithms for probabilistic first-order logic frameworks such as Markov logic networks (MLNs) have received significant attention in recent years. These algorithms use so called lifting rules to identify symmetries in the first-order representation and reduce the inference problem over a large probabilistic model to an inference problem(More)
This paper studies a new framework for learning a predictor in the presence of multiple kernel functions where the learner selects or extracts several kernel functions from potentially complex families and finds an accurate predictor defined in terms of these functions. We present an algorithm , Voted Kernel Regularization, that provides the flexibility of(More)
We propose a novel GPU variant for genetic algorithm that uses constant memory effectively to store and share the elite population across different blocks to obtain faster convergence rates. We compare our algorithm to a previous work which does not use constant memory, and show that using constant memory significantly improves the quality of solution for a(More)
Support Vector Machines (SVMs) are state-of-the-art algorithms for classification in machine learning. However, the SVM formulation does not directly seek to find sparse solutions. In this work, we propose an alternate formulation that explicitly imposes sparsity. We show that the proposed technique is related to the standard SVM formulation and therefore(More)
— This paper gives a glimpse of basic quantum computing concepts, which is an active area of research in today's world. The discussion starts with difference between a quantum computer and a classical one at the bit level, register level, logic gates level and at circuit level. Then, some of the important quantum phenomena such as quantum entanglement and(More)
Optimization problems are ubiquitous in machine learning, while MATLAB provides excellent implementations of various most commonly used optimization algorithms. In this work, we comapre the MATLAB implementations of various algorithms for optimization problems in machine learning.
Disclaimer: This report is submitted to NYU's Computer Science Department for the sole purpose of assigning an internship grade. The information remains confidential and proprietary to the company. Abstract We consider the problem of captioning videos in the wild using deep learning techniques. The aim was to improve over the existing state-of-the-art(More)
  • 1