A collection of machine learning notebooks and deep learning models.
Kaggle: https://www.kaggle.com/evelynchin
Table of Contents
- ASHRAE-energy-predictor -- uses a LightGBM to predict building energy usage in chilled water, electric, hot water, and steam meters
- house-prices-advanced-regression -- implements ensembles and stacked models to predict house prices based on tens of features
- nlp-disaster-tweets -- uses an LSTM to classify whether or not a tweet is about a real disaster or not
- cat-image-recognition -- a neural network that recognizes cat pictures from non cat pictures
- dinosaur-name-generator-rnn -- a character level language model that generates new dinosaur names, as well as a shakespeare poem generator :)
- emojify -- chooses an emoji to go along with an inputted sentence using a 2 layer LSTM model and embedding matrices
- emotion-detection-keras -- a neural network built through Keras that detects 'happy' vs. 'not happy' faces
- face-verification -- implements face verification and recognition through building the triplet loss function and loading a pretrained Inception model for encoding
- jazz-improvisation-lstm -- uses an LSTM network to generate a jazz solo improvisation through Keras and a 78-value musical encoding system
- mnist-numbers -- a neural network that classifies handwritten digits
- SIGNS-tf-cnn -- a convolutional neural network that classifies the numbers 0 to 5 of the SIGNS dataset
- traffic-signs -- a neural network that classifies and recognizes traffic signs
- art-generation-nst -- implements the neural style transfer algorithm to generate novel artistic images
- car-detection-yolo -- uses the YOLO algorithm to detect cars and objects on the road, setting up score-thresholding and non-max suppression
- cnn-setup -- sets up the fundamental functions and building blocks that are required of CNNs (convolution, forward pass, forward and backward pooling, etc)
- comparing-initializations -- a comparison of zero, random, and He initialization of parameters
- comparing-optimizations -- a comparison of batch gradient descent, stochastic gradient descent, and mini-batch gradient descent
- comparing-regularization -- a comparison of L2 regularization, dropout, and no regularization
- deep-neural-network-setup -- sets up the fundamental functions and processes that are required of neural networks (forward propagation, backward propagation, initializing parameters, etc)
- gradient-check -- implementing the process of gradient check to ensure back propagation works effectively
- linear-regression -- a simple implementation of linear regression, and my first introduction to machine learning
- lyft-data-analysis -- an exploratory data analysis of image and LiDAR data taken from a car
- neural-machine-translation-with-attention -- builds an NMT model that translates human-readable dates into machine-readable dates using the attention model
- operations-on-word-vectors -- implementing cosine similarity, solving analogy problems, and modifying word embeddings to reduce their gender bias
- resnet-setup -- sets up the identity and convolutional blocks of resnet, and builds a 50-layer ResNet through Keras
- rnn-setup -- sets up individual RNN and LSTM cells, and implements the forward propagation process of a recurrent neural network and an LSTM
- tensorflow -- an introduction to tensorflow and its usage of variables, constants, and sessions (essentially a tensorflow tutorial)
- trigger-word-detection -- synthesizes and processes audio recordings to create train/dev datasets and trains a trigger word detection model to make predictions