Ayush Adhikari

My personal website

Ayush Adhikari

 > 

Search
Monday, May 19, 2025

Found 3 result(s) for "Pandas"! Click on the links for more details

Japanese-To-English Translator

This is a translator program I built using Python that will take in a sentence in Japanese and translate the parts of the sentence into English and return an English translation. The original senetence is first split into various parts of speech using Japanese particles, which determines the transitivity of the verb if present and if a copula is present. Each part of speech is further broken down depending on the characters, length, conjugation, pairing, and use. This program does utilize the googletrans library as making a translator is a near impossible task, especially without the use of machine learning. The main feature of the program is to take the individual parts of the sentence, translate it using a csv with common words in Japanese, match it to a part of speech in English, and return that to the user. To learn more, click on the github icon next to the title of the project.

Data Augmentation Using Generative Adversarial Networks

This was a team project for my CS345 class, where I worked with a teammate to research and use GANs as a way to improve the accuracy of classifiers. We first researched neural networks and the workings of GANs to understand how the model functions. Then, we used TensorFlow as the framework to implement the GAN model to generate data that could be augmented to the test set. We trained the generator on various splits of the original training data and noted how effective the augmentation was. This was compared to the baseline accuracy of just the original data on a SVC classifier to train and test. The use case for this was to improve the data on which to test as a mean of improving the accuracy of the classifier with better data produced by the GAN.

CS345: Machine Learning Foundations and Practice

Introduced machine learning algorithms and tools for predictive modeling presented using case studies that inform their use in real-world applications. Utilized python, numpy, pandas, matplotlib, R, etc. for data analysis and predictions.

Nothing to show!
×

This is a feature mainly used to demonstrate the use of the handwriting app built using a canvas and CNN. You can use the search bar to search the site for content or to navigate the site by typing in a word such as home, which will take you to the home page. Below the search bar is a canvas where you can draw letters. This is implemented using a digit recognition CNN. As the accuracy is around 89 percent, it might not always produce an accuracte prediction, for example between o and 0. You can use the three button next to the search bar to delete the last character, clear the search bar, and finally search the site.