Related to:
Table - Single models with options
Detailed Breakdown of Popular Models and Architectures
Autoencoders
| Category | Details |
|---|---|
| Basic Components | Encoder, Decoder, Latent Space |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Reconstruction Loss |
| Model’s Parameters | Weights (Changeable), Latent Dimensions (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Reconstruction Accuracy |
| Model’s Hyperparameters | Learning Rate, Latent Dimension Size (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Lower Reconstruction Error |
| Basic Components of Hyperparameter’s Productivity | Effective Latent Space Size, Training Convergence Rate |
CNN (Convolutional Neural Networks)
| Category | Details |
|---|---|
| Basic Components | Convolution Layers, Pooling, Fully Connected Layers |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Accuracy, Precision, Recall, F1-Score |
| Model’s Parameters | Filter Weights (Changeable), Input Channels (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Detection Accuracy |
| Model’s Hyperparameters | Kernel Size, Stride, Number of Filters (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Higher Feature Extraction Quality |
| Basic Components of Hyperparameter’s Productivity | Filter Efficiency, Computational Cost |
RNN (Recurrent Neural Networks)
| Category | Details |
|---|---|
| Basic Components | Recurrent Layers, Activation Functions |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Perplexity, Accuracy, BLEU Score (NLP) |
| Model’s Parameters | Hidden State (Changeable), Sequence Length (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Temporal Pattern Capture Efficiency |
| Model’s Hyperparameters | Learning Rate, Hidden State Size (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Sequence Learning Performance |
| Basic Components of Hyperparameter’s Productivity | Effective Sequence Memory Size |
LSTM (Long Short-Term Memory Networks)
| Category | Details |
|---|---|
| Basic Components | LSTM Cells (Input, Forget, Output Gates) |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Perplexity, Accuracy (Time-Series, NLP) |
| Model’s Parameters | Cell Weights (Changeable), Memory Cell (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Long-Term Dependency Capture Efficiency |
| Model’s Hyperparameters | Learning Rate, Number of Layers (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Retention of Long-Term Dependencies |
| Basic Components of Hyperparameter’s Productivity | Sequence Retention and Gradient Stability |
GNN (Graph Neural Networks)
| Category | Details |
|---|---|
| Basic Components | Node Embeddings, Edge Features, Graph Convolutions |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Node Classification Accuracy, Link Prediction |
| Model’s Parameters | Edge Weights (Changeable), Node Attributes (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Graph Feature Capture Efficiency |
| Model’s Hyperparameters | Number of Layers, Embedding Size (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Graph-Level Feature Generalization |
| Basic Components of Hyperparameter’s Productivity | Graph Topology Learning |
BERT (Bidirectional Encoder Representations from Transformers)
| Category | Details |
|---|---|
| Basic Components | Encoder, Multi-Head Attention, Feedforward Layers |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | F1-Score, Exact Match (QA), Perplexity |
| Model’s Parameters | Token Embeddings (Changeable), Vocabulary (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Contextual Understanding Quality |
| Model’s Hyperparameters | Learning Rate, Batch Size, Sequence Length (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Contextual Embedding Accuracy |
| Basic Components of Hyperparameter’s Productivity | Attention Mechanism, Positional Encoding |
BART (Bidirectional and Auto-Regressive Transformers)
| Category | Details |
|---|---|
| Basic Components | Encoder-Decoder, Multi-Head Attention, Feedforward |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Rouge Score, BLEU Score |
| Model’s Parameters | Attention Weights (Changeable), Vocabulary (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Summarization and Translation Accuracy |
| Model’s Hyperparameters | Learning Rate, Number of Heads (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Text Generation Quality |
| Basic Components of Hyperparameter’s Productivity | Encoder-Decoder Consistency |
T5 (Text-to-Text Transfer Transformer)
| Category | Details |
|---|---|
| Basic Components | Encoder-Decoder, Attention Mechanisms, Feedforward |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Rouge Score, BLEU Score |
| Model’s Parameters | Token Embeddings (Changeable), Vocabulary (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Text-to-Text Conversion Accuracy |
| Model’s Hyperparameters | Sequence Length, Beam Width (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Text Generation Coherence |
| Basic Components of Hyperparameter’s Productivity | Attention Span, Latent Representation Quality |
LLAMA
| Category | Details |
|---|---|
| Basic Components | Transformer Layers, Feedforward Layers, Attention |
| Open-Source/ Forbidden | Restricted for Modifications |
| Criteria of Measuring Productivity | F1-Score, Rouge Score |
| Model’s Parameters | Attention Weights (Changeable), Vocabulary (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Latent Representation Consistency |
| Model’s Hyperparameters | Number of Layers, Head Size (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Layer-to-Layer Weight Propagation |
| Basic Components of Hyperparameter’s Productivity | Transformer Block Efficiency |
GPT (Generative Pre-trained Transformer)
| Category | Details |
|---|---|
| Basic Components | Transformer Decoder, Feedforward Layers, Attention |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Perplexity, BLEU Score |
| Model’s Parameters | Attention Weights (Changeable), Vocabulary (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Generative Text Coherence |
| Model’s Hyperparameters | Learning Rate, Model Depth, Token Limit (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Generative Text Quality |
| Basic Components of Hyperparameter’s Productivity | Token Context Understanding |
ViT (Vision Transformer)
| Category | Details |
|---|---|
| Basic Components | Patch Embedding, Transformer Layers, Attention |
| Open-Source/ Forbidden | Open-Source |
| Criteria of Measuring Productivity | Accuracy, Precision, Recall, F1-Score |
| Model’s Parameters | Patch Embeddings (Changeable), Image Size (Fixed) |
| Criteria of Measuring Parameter’s Productivity | Visual Feature Generalization |
| Model’s Hyperparameters | Patch Size, Attention Heads (Changeable) |
| Criteria of Measuring Hyperparameter’s Productivity | Patch Extraction Accuracy, Attention Span |
| Basic Components of Hyperparameter’s Productivity | Image Feature Learning Efficiency |