Skip to content

Latest commit

 

History

History
107 lines (86 loc) · 21.3 KB

README.md

File metadata and controls

107 lines (86 loc) · 21.3 KB

Anything_From_Scratch

A small repo that we want to give some value to the community about implementing artificial intelligence algorithms, as well as training our ability to code and understand algorithms better.

Update

With some models I research (I will note the star next to the model name), there will be small changes in the file layout. An algorithm that I learn will have a small readme.md section and more code to try more problems.

Contents

Machine Learning

Title Source Category/ Tag Description
Linear Regression LinearRegression Machine Learning
Logistic Regresssion Logistic Regresssion Machine Learning
K Means Clustering K Mean Clustering Machine Learning
K Near Neighbor K Near Neighbor Machine Learning
Decision Trees Decision Trees Machine Learning
Naive Bayes Naive Bayes
PCA PCA Machine Learning, Dimension Reduce
SVM SVM Machine Learning, Dimension Reduce

Computer Vision

Title Source Category/ Tag Description
LeNet LeNet Image Classification, CNN architechture
AlexNet AlexNet Image Classification, CNN architechture
VGG VGG Image Classification, CNN architechture
ResNet ResNet Image Classification, CNN architechture
MobileNet MobileNet Image Classification, CNN architechture
Effecient Net Effecient Net Image Classification, CNN architechture
Vision in Transformer ViT Image Classification, CNN architechture
SORT SORT Object Tracking
SORTwithYolov8 SORTwithYolov8 Object Tracking

NLP

Title Source Category/ Tag Description Slide
Dense Vectors Dense Vectors NLP
Words2Vec Words2Vec NLP
DeepRecurrentNeuralNetworks DRNN NLP
GRU GRU NLP
LSTM LSTM NLP
Attention Attention
Self-Attention Self-Attention
Multihead Attention MultiheadAttention
Positional Encoding Positional Encoding
Transformers Transformers Transformer
BERT BERT BERT
QnA with BERT QnABert
GPT v1 GPTv1
GPT v2 GPTv2
GPT v3 GPTv2
DistillBERT DistillBERT
T5 T5
LoRA LoRA
BLIP-2* BLIP-2 LVLMs(Vision-language) A appoach to connect Vision and LLMs together. Demo on Image Captioning and Chatbot
QLoRA* QLoRA Optimization in Memory A approach to optimise for LLMs. Demo in random model. QLoRA
TaPas* TaPas Question and Answer TaPas use BERT model as a base for QnA task with table datastructure. TaPas
Rotary Positions Embedding (RoPE) RoPE Position Embedding Rotary Position Embedding (RoPE) is a novel method for encoding relative positional information into transformer-based models. Unlike traditional position encoding methods that add positional information to embeddings, RoPE encodes relative positions by rotating word embeddings in a multi-dimensional space using a rotation matrix.

VLMs

Title Source Category/ Tag Description Slide
MolMo (Multimodel Open Languale Model) Tự tìm trong folder trước ạ.

Generative Image Model

Title Source Category/ Tag Description
GAN GAN GM ---
DCGAN DCGAN GM ---
Diffusion Diffusion GM ---

Audio

Title Source Category/ Tag Description

Pytorch Basics

Title Source Category/ Tag Description
Tensor Basic TensorBasics
Full Neural Network FNN
CNN CNN
LSTM LSTM
Load and Save Model LoadSave
Bidirectional Bidirectional
CustomImageDataset CustomImageDataset

Other

Title Source Category/ Tag Description