ML-learning

Samples of some ML algorithms and visualisations

STS (Semantic Textual Similarity)

• cleaning data;
• preprocessing data for transformers;
• creating embeddings with paraphrase-multilingual-MiniLM-L12-v2 Transformer;
• creating embeddings with universal-sentence-encoder-multilingual Transformer;
• comparing embeddings Cosine similarity.

Efficent Apriori and FP-Growth - Frequent itemset mining algorithms

• prepocessing data;
• creating frequent itemset and rules;
• visualising Efficent Apriori rules with PyARMViz library;
• visualize and output FP-tree with Graphviz;
• testing the running time of the algorithms.

KNN, SVM, RF algorithms + visualising with t-SNE and UMAP

• prepocessing data;
• normalizing data with StandardScaler;
• balancing train data with SMOTE;
• GridSearch hyperparameters and comparing results;
• visualising models with t-SNE and UMAP;
• checking CatBoost algorithm with the same data;
• looking feature importance (FI) with CatBoost and SHAP.

Rosenbrock function Optimization (Newton, GD, PSO, GA)

• plotting Ronsenbrock function 3D and 2D;
• implementing Newton's method;
• implementing Gradient Descent;
• implementing Particle Swarm Optimization (PSO) algorithm;
• implementing Genetic Algorithm algorithm (GA);
• looking at Law of large numbers and Central limit theorem;
• comparing GA, PSO and Gradient Descent;
• animating GD and PSO algorithms.

K-means, DBSCAN and FCM clusterizations

• creating clusters with 10 centers;
• K-Means clusterization with 2-10 clusters;
• DBSCAN clusterization with different eps and min_samples;
• Fuzzy-C-Means clusterization with 2-10 clusters;
• comparing best clusterizations.

Recurrent Neural Networks (Time Series) - LTSM, GRU, Simple_RNN

• prepocessing data;
• normalizing data with MinMax scaler;
• generating time series;
• implementing LSTM model;
• implementing GRU model;
• implementing SimpleRNN model;
• comparing results on CPU and GPU learning (Google Collab).