🧠 Unified AI Framework

A powerful and extensible deep learning + machine learning + NLP framework built in Python. Designed to be educational, modular, and production-ready.

🚀 Features

Core Framework

• Tensor abstraction using NumPy with automatic differentiation support
• Layer classes: Dense, Conv2D, BatchNorm, Dropout, Embedding, Attention, etc.
• Model class (Keras-like): supports compile(), fit(), evaluate(), predict()
• Optimizers: SGD, Adam, RMSProp with momentum and weight decay
• Callbacks: EarlyStopping, ModelCheckpoint, LearningRateScheduler
• Loss functions: CrossEntropy, MSE, BCE, Focal Loss
• Metrics: Accuracy, Precision, Recall, F1, BLEU, IoU

Machine Learning (from scratch)

• K-Nearest Neighbors (KNN)
• Support Vector Machine (SVM)
• Decision Tree with pruning
• Naive Bayes (Gaussian, Multinomial, Bernoulli)
• Linear/Logistic Regression with regularization

Deep Learning Models

• GAN: Deep Convolutional GAN (DCGAN) with Wasserstein loss
• YOLO: Tiny YOLO for object detection with NMS and anchor boxes
• Transformer: GPT-like transformer with multi-head attention

NLP Pipeline

• Tokenizer: Byte Pair Encoding (BPE) and WordPiece
• Embeddings: Word2Vec, GloVe with pre-trained support
• Language Model: Transformer-based causal language modeling
• Preprocessing: Text cleaning, stemming, lemmatization

📦 Installation

Or install from source:

git clone https://github.com/Drxmukesh/unified-ai-library.git
cd unified-ai-library
pip install -e .

🏃‍♂️ Quick Start

Deep Learning Example

from unified_ai.core import Model, Dense, Dropout
from unified_ai.utils.losses import CrossEntropyLoss
from unified_ai.core.optimizer import Adam

# Create a simple neural network
model = Model([
    Dense(128, activation='relu', input_shape=(784,)),
    Dropout(0.2),
    Dense(64, activation='relu'),
    Dense(10, activation='softmax')
])

# Compile the model
model.compile(
    optimizer=Adam(learning_rate=0.001),
    loss=CrossEntropyLoss(),
    metrics=['accuracy']
)

# Train the model
history = model.fit(X_train, y_train, 
                   validation_data=(X_val, y_val),
                   epochs=50, batch_size=32)

Machine Learning Example

from unified_ai.ml import SVM, DecisionTree
from unified_ai.utils.metrics import accuracy_score

# Support Vector Machine
svm = SVM(kernel='rbf', C=1.0)
svm.fit(X_train, y_train)
y_pred = svm.predict(X_test)
print(f"SVM Accuracy: {accuracy_score(y_test, y_pred)}")

# Decision Tree
dt = DecisionTree(max_depth=10, min_samples_split=5)
dt.fit(X_train, y_train)
y_pred = dt.predict(X_test)
print(f"Decision Tree Accuracy: {accuracy_score(y_test, y_pred)}")

NLP Example

from unified_ai.nlp import BPETokenizer, Word2Vec
from unified_ai.models import Transformer

# Tokenization
tokenizer = BPETokenizer(vocab_size=10000)
tokenizer.fit(corpus)
tokens = tokenizer.encode("Hello, world!")

# Word embeddings
w2v = Word2Vec(vector_size=300, window=5, min_count=1)
w2v.fit(sentences)
embedding = w2v.get_vector("hello")

# Transformer language model
model = Transformer(
    vocab_size=10000,
    d_model=512,
    n_heads=8,
    n_layers=6,
    max_seq_len=1024
)

🏗️ Architecture

unified_ai/
├── core/           # Core framework components
├── callbacks/      # Training callbacks
├── ml/            # Traditional ML algorithms
├── models/        # Deep learning architectures
├── nlp/           # NLP components
├── utils/         # Utilities and helpers
├── examples/      # Example scripts
└── tests/         # Unit tests

🧪 Examples

Check out the examples/ directory for complete training scripts:

• train_gan.py - Train a DCGAN on MNIST
• train_yolo.py - Object detection with YOLO
• train_transformer.py - Language modeling with Transformer
• train_ml_models.py - Traditional ML algorithms comparison
• nlp_pipeline_example.py - Complete NLP pipeline

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Inspired by TensorFlow/Keras, PyTorch, and scikit-learn
• Built for educational purposes and research
• Community-driven development