mihaelagrigore

Data-Exploration-Visualisation-Feature-Engineering-Pandas-Numpy-Seaborn

In this notebook I go thoroughly through the normal Data Wrangling process performed on any new and potentially messy data. It includes analysis of missing data, uni and bivariate analysis, removal of data that is not useful and some data engineering.

Deep-Learning-CNN-with-PyTorch

Building a custom architecture Convolutional Neural Network in PyTorch and using the resulting model to classify the Fashion MNIST dataset.

Deep-Learning-Face-Recognition-Tensorflow-Keras

Facial Recognition with Keras, FaceNet and Inception model using Siamese Networks and transfer learning.

Deep-Learning-Keras-Tensorflow-Resnet50-implementation

Building a 50-layer ResNet model from scratch using Tensorflow and Keras. Training it first on CPU (very slow), then on Kaggle GPU (for a significant improvement in speed).

Exploratory-Data-Analysis

Introduction to Exploratory Data Analysis (EDA). Made for beginners. Shows how to look at data to uncover patterns.

FineTune-DistilBERT

Huggingface transformers: Finetuning DistilBERT on a toxic comment binary classification task.

Gutstuck

Stuck in a weak human body. Help the microbiota recover and wander out these gutsy meanders !

NLP-for-sentiment-analysis-of-tweets

Sentiment analysis performed on tweets. Comparison of performance of a few popular libraries: TextBlob, VADER and Flair. Also includes some pretty plots using Keppler.

Scraping-historical-tweets-without-Twitter-API

The tool I'm using here for scraping tweets: snscrape. Snscrape is a popular tool with social scientists for Tweets collection, at least in 2021. Apparently, it bypasses several limitations of the Twitter API. The prettiest thing is that you don't need Twitter developer account credentials (like you do with Tweepy, for example)

Time-series-exploration-Crypto-Price

I wanted to see a notebook / tutorial that would take me through the basics of working with time series. Most notebooks I found were not very rigorous. So I wrote my own. This is: 1. a good starting point for understanding time series data and how it differs from problems with other type of tabular data 2. a cookbook to be used for exploration when starting to work with a new dataset

mihaelagrigore/Deep-Learning-Keras-Tensorflow-Resnet50-implementation

Building a 50-layer ResNet model from scratch using Tensorflow and Keras. Training it first on CPU (very slow), then on Kaggle GPU (for a significant improvement in speed).

mihaelagrigore/Scraping-historical-tweets-without-Twitter-API

The tool I'm using here for scraping tweets: snscrape. Snscrape is a popular tool with social scientists for Tweets collection, at least in 2021. Apparently, it bypasses several limitations of the Twitter API. The prettiest thing is that you don't need Twitter developer account credentials (like you do with Tweepy, for example)

mihaelagrigore/Gutstuck

Stuck in a weak human body. Help the microbiota recover and wander out these gutsy meanders !

mihaelagrigore/Data-Exploration-Visualisation-Feature-Engineering-Pandas-Numpy-Seaborn

In this notebook I go thoroughly through the normal Data Wrangling process performed on any new and potentially messy data. It includes analysis of missing data, uni and bivariate analysis, removal of data that is not useful and some data engineering.

mihaelagrigore/Deep-Learning-CNN-with-PyTorch

Building a custom architecture Convolutional Neural Network in PyTorch and using the resulting model to classify the Fashion MNIST dataset.

mihaelagrigore/Deep-Learning-Face-Recognition-Tensorflow-Keras

Facial Recognition with Keras, FaceNet and Inception model using Siamese Networks and transfer learning.

mihaelagrigore/Exploratory-Data-Analysis

Introduction to Exploratory Data Analysis (EDA). Made for beginners. Shows how to look at data to uncover patterns.

mihaelagrigore/FineTune-DistilBERT

Huggingface transformers: Finetuning DistilBERT on a toxic comment binary classification task.

mihaelagrigore/NLP-for-sentiment-analysis-of-tweets

Sentiment analysis performed on tweets. Comparison of performance of a few popular libraries: TextBlob, VADER and Flair. Also includes some pretty plots using Keppler.

mihaelagrigore/Time-series-exploration-Crypto-Price

I wanted to see a notebook / tutorial that would take me through the basics of working with time series. Most notebooks I found were not very rigorous. So I wrote my own. This is: 1. a good starting point for understanding time series data and how it differs from problems with other type of tabular data 2. a cookbook to be used for exploration when starting to work with a new dataset

mihaelagrigore/Network-Analysis

A quick demo of network analysis and network visualisation.

mihaelagrigore/NLP-tweets-sentiment-analysis-topic-modeling-LDA

Calculate topic coherence for topic models. Discover the semantic structure of texts by examining statistical co-occurrence patterns within a corpus of training documents (tweets) using Latent Dirichlet Allocation.

mihaelagrigore/rag-from-scratch

mihaelagrigore/real_time_sign_language_detection_zoom

Real Time object detection using browser deployed Tensorflow.js model

mihaelagrigore/Socia-Media-Text-Scraping-Twitter-Developer-API-for-Academics

Social media scraping and NLP for research | Scraping unlimited number of historical tweets using the Academic Track of Twitter Developer API. Exploring the metadata obtained. Text analysis using NLP.

mihaelagrigore/time-series-prediction-prophet-arima

Time series prediction (demand forecast) on a Kaggle competition dataset using smoothing methods, Prophet (+Neural Prophet), ARMA (and several of its versions).