Tirth Patel

Variational Auto-Encoders and Bayesian Optimization.

• Facial composites are widely used in forensics to generate images of suspects. Since the victim or witness usually isn’t good at drawing, computer-aided generation is applied to reconstruct the face attacker. One of the most commonly used techniques is evolutionary systems that compose the final face from many predefined parts.

• In this project, I implement an app for creating a facial composite that will be able to construct desired faces without explicitly providing databases of templates. I apply Variational Autoencoders, Gaussian processes, and Bayesian Optimization for this task.

• Skills Applied: Bayesian Modelling, Bayesian Optimization, Gaussian Processes, Variational Auto Encoder, Variational Inference, Convolutional Neural Networks.

• Frameworks Used: TensorFlow 1.x, NumPy, SciPy, matplotlib, GPy, GPyOpt
• See the Project Here

LSTMs and Convolutional Neural Networks!

• I have used TensorFlow’s pretrained InceptionNet as a encoder network and dynamic stacked LSTM as a decoder network. All the images provided as input are converted to vectors of a fixed length and this vectors are a input to the one-to-many LSTM that outputs a english caption for the image.
• I have used finetuning using Keras for the task for about 30 epochs. It produces reasonable captions while testing on internet images. You can also test an image by uploading it to the internet and running the last couple of cells of the notebook.

• Skills Applied: Convolutional Neural Networks, Long Short Term Memory, NLP, Image Preprocessing

• Frameworks used: TensorFlow 1.x, NumPy, SciPy, scikit-learn
• See the Project Here

GANs and VAEs are just awesome!

• I trained a Variational Auto Encoder from scratch for 100 epochs on a training set of 60,000 images and 3 and 30 dimensional latent space. The generator generated very clear fake handwritten images.

• Skills Applied: Convolutional Neural Networks, Image Preprocessing, Bayesian Modelling, Variational Auto Encoders

• Frameworks used: TensorFlow 1.x, NumPy, SciPy, matplotlib
• See the Project Here

This article explores the famous graph traversals, namely, Depth First Search (DFS), Breadth First Search (BFS), Uniform Cost Search (Dijkstra’s algorithm), A-Star Search, Weighed A-Star Search and Learning Real Time A-Star Search algorithms with interesting visualizations.

• I created and explained animations for various search algorithms in Python.
• Published in Analytics Vidhya
• See Article
• See GitHub Repository

I write about sciences that interest me.

• I write multiple articles on Machine Learning, AI, Physics, and Mathematics on my website. I have been maintaining it since January 2020
• Visit my Website
• See GitHub Repository

A talk on the impact of Data Science on Astronomy.

• Delivered a 6 hour talk on how big data is used in astronomy to perform data analytics and machine learning methods to predict redshifts.
• See GitHub Repository

In this article, I have implemented a fully vectorized code for Artificial Neural Network with Dropout and L2 Regularization.

• In this article, I have implemented a fully vectorized python code of an Artificial Neural Network tested on multiple datasets. Further, dropout and L2 regularization techniques are implemented and explained in detail.
• Got 200+ claps (appreciation on medium).
• Published in Towards Data Science.
• See Article
• See GitHub Repository