Skip to content
franklinic edited this page Jan 19, 2026 · 2 revisions

AiDotNet Wiki

Welcome to the official AiDotNet documentation wiki! AiDotNet is the most comprehensive AI/ML framework for .NET, featuring 100+ neural network architectures, 106+ classical ML algorithms, and much more.

Quick Links

Resource Description
Getting Started Install AiDotNet and build your first model
Installation Guide Detailed installation instructions
Interactive Playground Try AiDotNet in your browser
API Reference Auto-generated API documentation
GitHub Repository Source code and issues

Feature Overview

Neural Networks (100+ Architectures)

  • Feedforward: Dense, MLP, Residual connections
  • Convolutional: CNN, ResNet, VGG, EfficientNet, MobileNet
  • Recurrent: RNN, LSTM, GRU, Bidirectional
  • Transformers: Attention, Multi-head attention, Vision Transformer
  • Generative: GAN, VAE, Diffusion models
  • Graph: GNN, GAT, GraphSAGE
  • Specialized: Capsule networks, NeRF, Neural ODE

Learn more about Neural Networks

Classical Machine Learning (106+ Algorithms)

  • Classification (28): SVM, Random Forest, Gradient Boosting, Naive Bayes, KNN, Decision Trees
  • Regression (41): Linear, Ridge, Lasso, ElasticNet, SVR, Gaussian Process
  • Clustering (20+): K-Means, DBSCAN, HDBSCAN, Spectral, Hierarchical
  • Dimensionality Reduction: PCA, t-SNE, UMAP, LDA

Learn more about Classical ML

Computer Vision (50+ Models)

  • Object Detection: YOLO v8-11, DETR, Faster R-CNN
  • Segmentation: Mask R-CNN, U-Net, SAM
  • Classification: ResNet, EfficientNet, Vision Transformer
  • OCR: Text detection and recognition

Learn more about Computer Vision

Audio Processing (90+ Models)

  • Speech Recognition: Whisper, Wav2Vec2
  • Text-to-Speech: Multiple TTS engines
  • Audio Classification: Sound event detection
  • Music Generation: AudioLDM, MusicGen

Learn more about Audio Processing

Natural Language Processing

  • Text Classification: Sentiment, topic classification
  • Named Entity Recognition: NER models
  • Embeddings: Sentence transformers, word embeddings
  • Tokenization: BPE, WordPiece, SentencePiece

Learn more about NLP

RAG & LLMs

  • Vector Stores: In-memory, persistent stores
  • Retrievers: Dense, sparse, hybrid retrieval
  • Rerankers: Cross-encoder reranking
  • Chunking: Semantic, recursive chunking

Learn more about RAG

LLM Fine-tuning (37+ LoRA Variants)

  • Standard LoRA: Low-rank adaptation
  • QLoRA: Quantized LoRA
  • DoRA: Weight-decomposed LoRA
  • AdaLoRA: Adaptive LoRA

Learn more about Fine-tuning

Reinforcement Learning (80+ Agents)

  • Value-based: DQN, Double DQN, Dueling DQN
  • Policy Gradient: PPO, A2C, A3C, TRPO
  • Actor-Critic: SAC, TD3, DDPG
  • Model-based: World models, MuZero
  • Multi-Agent: MADDPG, QMIX

Learn more about Reinforcement Learning

Distributed Training

  • Data Parallel: DDP, DistributedDataParallel
  • Model Parallel: Pipeline, Tensor parallelism
  • ZeRO: ZeRO-1, ZeRO-2, ZeRO-3
  • Cloud: vast.ai, RunPod integration

Learn more about Distributed Training

Tutorials by Task

Task Tutorial
Binary Classification Sentiment Analysis
Multi-class Classification Iris Classification
Regression House Price Prediction
Image Classification CIFAR-10 with CNN
Object Detection YOLO Object Detection
Text Generation RAG Chatbot
Fine-tuning LoRA Fine-tuning
Reinforcement Learning CartPole with DQN

Why AiDotNet?

Feature AiDotNet ML.NET TorchSharp TensorFlow.NET
Neural Network Architectures 100+ ~10 Via PyTorch Via TensorFlow
Classical ML Algorithms 106+ ~30 Limited Limited
Native .NET Yes Yes Wrapper Wrapper
HuggingFace Integration Yes No Partial No
LoRA/PEFT Support 37+ variants No Manual No
Distributed Training Full stack Limited Via PyTorch Via TensorFlow

Getting Help

Contributing

Want to contribute? See our Contributing Guide.

License

AiDotNet is licensed under the MIT License. See LICENSE for details.

Clone this wiki locally