10 GitHub Repositories for Advanced Machine Learning Projects

Image by Author

GitHub really is a hub for learning many things, machine learning being only one of them. It’s a rich repository source where you can get lost in machine learning projects. Literally lost, and that’s not a good thing.

Here’s a plan to get you out of the woods. First, I’ll define what advanced machine learning actually is. Then, I’ll browse GitHub and find some good repositories for advanced ML projects.

 

What Does Advanced ML Encompass?

It would be nice if there were a standardized definition of advanced machine learning. There isn’t, but from my experience, these eleven topics are what is generally considered advanced.

 

 

1. Deep Learning

Deep learning (DL) uses multi-layered (deep) neural networks to simulate the functioning of a human brain when learning. Some examples of typical DL architectures are Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Generative Adversarial Networks (GANs).

 

2. Reinforcement Learning

Reinforcement learning (RL) refers to training AI agents to make decisions by interacting with a dynamic environment and maximizing cumulative reward. This approach is commonly used in autonomous systems, game AI, and optimization tasks.

 

3. Transfer Learning

Transfer learning is an approach popular in deep learning and involves taking a pre-trained model and applying it to a different but related problem, e.g., using a pre-trained image recognition model on a new dataset of medical images.

 

4. Ensemble Learning

Ensemble learning combines multiple models and their predictions to construct more accurate predictions. In doing so, you can employ common techniques, such as boosting (e.g., XGBoost, AdaBoost), bagging (e.g., random forest), and stacking.

 

5. Natural Language Processing

Natural language processing (NLP) is a subset of AI that involves understanding and generating human (natural) language. Some techniques used are transformers (e.g., BERT, GPT), named entity recognition (NER), text generation, and summarization.

 

6. Self-Supervised Learning

Self-supervised learning is an ML paradigm based on neural networks where the models are trained on unlabeled data, and the model creates the labels from the data itself.

 

7. Bayesian Methods

Bayesian machine learning is based on Bayes’s theorem to handle uncertainties in predictions. Common applications include Bayesian neural networks (BNN), Gaussian processes (GP), Bayesian optimization, Bayesian inference in hierarchical models, Markov chain Monte Carlo (MCMC) methods, Bayesian decision theory, Bayesian deep learning, etc.

 

8. Multimodal Machine Learning

Multimodal machine learning is a part of DL where learning is performed from different data modalities, such as text, images, and audio. Some examples of multimodal ML are image captioning and speech-driven animation.

 

9. Recommender Systems

Recommender systems are ML systems that learn from customer preference data and try to provide them with personalized suggestions, e.g., songs, artists, movies, and products. Advanced recommender systems employ collaborative filtering, content-based filtering, hybrid models, and DL techniques.

 

10. Meta-Learning

Meta-learning is an approach where models learn from other models’ outputs. It is used in scenarios where there’s minimal data and/or quick adaptability to a changing environment is required.

 

11. Time Series Analysis

The analysis of time series is a method of analyzing a sequence of data points, namely time series. Deep learning methods, such as RNN, long short-term memory (LSTM) RNNs, and attention mechanisms.

 

GitHub Repositories

Let’s now find ten GitHub repositories where you can find projects for practicing these ML topics.

 

1. gimseng/99-ML-Learning-Projects

Link: 99 ML Learning Projects Repository

Description: This repository currently contains ten ML projects, with the goal of reaching 99, hence the name. There are five projects I would consider advanced. First, there’s a project where you can learn Bagging and Boosting Ensemble Methods. Then, there’s a computer vision MNIST Handwriting Digit Recognition project. Next, there are two NLP projects, namely Sentiment Analysis and Text-Generation Neural Network Model (with LSTM). Finally, you can do the Naive Bayes Classification project.

Topics Learned: DL, Ensemble Methods, NLP, Recommender Systems, Bayesian methods

 

2. rohankrgupta/Orca-call-Classifier-Machine-learning

Link: Orca Call Classifier Repository 

Description: This is quite an unusual project that focuses on classifying orca calls. The project dataset consists of 240 mel-spectrograms, with each representing a 10-second audio of an orca call/no call (120 of each). Typically, you’d use CNNs, but due to the dataset being small, classification is performed using a random forest classifier to achieve better performance. The resulting model shows 88% accuracy.

In addition, this project involves analyzing time-dependent audio signals, for which you have to apply time-series analysis techniques.

Topics Learned: Ensemble Methods

 

3. Mehrab-Kalantari/Multi-Modal-House-Price-Estimation

Link: House Price Estimation From Visual and Textual Features Repository

Description: Another interesting project, this one focusing on estimating house prices. Typically, systems for automatic house price estimation rely only on textual information. This project takes another approach, and, along with textual data, visual features extracted from house photographs are used to estimate house prices. The dataset is comprised of 535 houses in California.

Along with classic ML algorithms, such as linear, polynomial, ridge, and decision tree regressions, you’ll also work with advanced ML models that use bagging and boosting methods. These are random forest regressor, support vector regressor, CatBoost regressor, and XGBoost regressor.

There’s also a part where you use a DL approach, i.e., multilayer perceptron (MLP) and CNNs.

Topics Learned: Deep Learning, Multimodal Machine Learning, Ensemble Learning

 

4. inboxpraveen/movie-recommendation-system

Link: Image Segmentation Repository

Description: This repository gives you an end-to-end pipeline for building a movie recommendation system. It uses a dataset from Kaggle and focuses on text-based feature extraction and similarity measurement to create a recommender model. The project will teach you several advanced ML techniques, such as count vectorizer (Bag of Words), cosine similarity, N-grams, and vector space model (VSM).

Topics Learned: Recommender Systems, NLP

 

5. souvikmajumder26/Land-Cover-Semantic-Segmentation-PyTorch

Link: Land Cover Semantic Segmentation Repository

Description: In this project, you will work on image segmentation, specifically semantic segmentation. The dataset from LandCover.ai is used to train U-Net, a type of CNN used specifically for image segmentation tasks. To improve learning efficiency, the model uses a pre-trained EfficientNet encoder.

Topics Learned: Deep Learning, Transfer Learning

 

6.ramyananth/Music-Recommender-System-using-ALS-Algorithm-with-Apache-Spark-and-Python

Link: Music Recommender System Repository

Description: This is another recommender system project, this time recommending music. It uses data from Audioscrobbler, which contains implicit ratings of tracks, i.e., the number of times a user played songs by an artist. In other words, the recommendation won’t be based on explicit ratings, e.g., the number of stars given to a song/artist by a user.

To handle this implicit feedback, the project employs the Alternating Least Squares (ALS) algorithm.

Topics Learned: Recommender Systems

 

7. antonio-f/Adversarial-Task

Link: Generative Adversarial Networks Repository

Description: This project is a solution to a test from Coursera’s Advanced Machine Learning - Intro to Deep Learning course. It builds a model that can generate believable images of human faces. It creates two neural networks; one is a generative adversarial network (GAN) that produces a face image, and the other is a usual convolutional network that takes the generated face image and tries to determine if it’s fake or not.

Topics Learned: Deep Learning, Self-Supervised Learning

 

8. firaja/flowers-classification

Link: Flowers Classification Repository

Description: In this project, you will build an ML model to classify flower images. The idea is to train the model on a small dataset (the 102 Category Flower Dataset) so that it can accurately classify flower images. With this, the possibility of model overfitting arises, which is attempted to circumvent by employing DL and transfer learning.

Topics Learned: Deep Learning, Transfer Learning

 

9. beimingliu/AdvancedMachineLearning

Link: Advanced Machine Learning Related Projects Repository

Description: This is a collection of projects with advanced ML topics in the projects such as Click-Through Rate Prediction (random forest), Spam Classification (Adabost, XGBoost), Neural Networks for MNIST Dataset (DL), Movie Review Classification (XGBoost), BBC Articles Recommendation (NLP), Movie Recommendation Systems (recommendation systems), Twitter Sentiment Analysis.

Topics Learned: Sentiment Analysis, NLP

 

10. mohammadmozafari/advanced-machine-learning

Link: Advanced Machine Learning Repository

Description: This repository consists of projects that implement several papers on advanced ML topics. These are the Transfer Learning Project, the Multi-Task Learning Project, the Black-Box Meta-Learning (SNAIL) Project, the Model Agnostic Meta-Learning (MAML), the Prototypical Networks Project, the Goal-Conditioned Reinforcement Learning and Hindsight Experience Replay (HER) Project, the Diversity is All You Need (DIAYN) Project, the Meta-Reinforcement Learning Project, the Gradient Episodic Memory (GEM) for Continual Learning

Topics Learned: Transfer Learning, Meta-Learning, Reinforcement Learning, Continual Learning

 

Conclusion

 

There you have it – ten GitHub repositories where you can practice advanced Machine Learning projects.

The topics range from time-series analysis, recommender systems, NLP, and meta-learning to Bayesian methods, self-supervised, ensemble, transfer, reinforcement, multimodal, and deep learning.

I think you’ll have a productive and enjoyable time doing these projects. Enjoy!

Nate Rosidi is a data scientist and in product strategy. He's also an adjunct professor teaching analytics, and is the founder of StrataScratch, a platform helping data scientists prepare for their interviews with real interview questions from top companies. Nate writes on the latest trends in the career market, gives interview advice, shares data science projects, and covers everything SQL.