matplotlab 0.1.13

Extended plotting and ML utilities library - Now with 10 ANN Labs (Autoencoders, RNN, LSTM)

Matplotlab

Extended plotting and machine learning utilities library for educational purposes.

A comprehensive Python library providing:

• Reinforcement Learning (RL) - Monte Carlo, TD Learning, Policy/Value Iteration, Dynamic Programming
• Artificial Neural Networks (ANN) - Deep Learning implementations with PyTorch and TensorFlow (10 Complete Labs!)
• Speech Processing (SP) - Audio analysis, MFCC extraction, vowel synthesis, formant analysis, dataset loading
• Visualization Tools - Enhanced plotting capabilities for ML workflows

NEW in v0.1.8: Complete ANN Labs 8, 9, 10 (Autoencoders, RNN, LSTM) - 40+ new functions!

Installation

# Install from PyPI
pip install matplotlab

# Or install from source
git clone https://github.com/Sohail-Creates/matplotlab.git
cd matplotlab
pip install -e .

Quick Start

Reinforcement Learning

from matplotlab import rl

# Create environment and find optimal policy
env = rl.create_frozenlake_env()
policy, V, iterations = rl.policy_iteration(env, gamma=0.99)
print(f"Converged in {iterations} iterations")

# Visualize results
rl.plot_value_heatmap(V)
rl.plot_grid_policy(policy)

# NEW: See complete lab workflow (Lab 1-6 + OEL)
rl.flowlab3()  # Shows complete Lab 3 code from import to visualization
rl.flowlab5()  # Shows complete Lab 5 (Policy Iteration) workflow

Artificial Neural Networks

from matplotlab import ann
import torch.nn as nn

# NEW: Lab 8 - Autoencoders
model = ann.UndercompleteAutoencoder()
train_loader, test_loader = ann.load_mnist_for_autoencoder()
losses = ann.train_undercomplete_autoencoder(model, train_loader, epochs=20)
ann.visualize_reconstructions(model, test_loader)

# NEW: Lab 9 - RNN for sentiment analysis
sentiment_model = ann.SentimentRNN(vocab_size=10000)
ann.train_sentiment_rnn(sentiment_model, epochs=10)

# NEW: Lab 10 - LSTM text generation
lstm = ann.NextWordLSTM(vocab_size=500)
text = ann.generate_text(lstm, "hello world", vocab, num_words=20)

# Classic MLP and CNN (Labs 1-7)
model = ann.create_mlp_model(input_size=10, hidden_sizes=[16, 8], output_size=1)
cnn_model = ann.create_fashion_cnn()

# See complete lab workflows
ann.flowlab8()   # Autoencoders workflow
ann.flowlab9()   # RNN workflow
ann.flowlab10()  # LSTM workflow

Speech Processing (NEW in v0.1.6!)

from matplotlab import sp

# Synthesize vowel sounds
audio, sr = sp.synthesize_vowel('A', f0=150, duration=1.0)
sp.play_audio(audio, sr)

# Analyze formants (OEL - important for teacher questions!)
formants = sp.identify_formants(audio, sr)
# Shows: Waveform + Spectrum + Spectrogram with F1, F2, F3 marked

# Extract MFCC features
mfcc = sp.extract_mfcc(audio, sr, n_mfcc=13)
# Or show manual implementation
mfcc_manual = sp.extract_mfcc_from_scratch(audio, sr)

# Load datasets with automatic preview
dataset = sp.load_audio_dataset('vowels.zip', sr=16000)
# Automatically shows: structure, waveforms, sample info, audio players

# Generate synthetic datasets
dataset = sp.generate_vowel_dataset(n_samples_per_vowel=100)

# Plot spectrograms and analyze pitch
sp.plot_mel_spectrogram(audio, sr)
sp.plot_pitch_histogram(audio, sr)

# Complete OEL workflow
sp.flowoel()  # Shows complete vowel synthesis code

Features

✅ Reinforcement Learning Module (44 functions)

• Environments: FrozenLake, Custom GridWorld
• Algorithms: Monte Carlo, TD Learning, Policy Iteration, Value Iteration
• MDP Utilities: State transitions, reward functions, probability computations
• Visualization: Heatmaps, policy arrows, convergence plots
• Lab Workflows (NEW): Complete code references for Labs 1-6 and OEL
  • flowlab1() through flowlab6() - Show full lab code workflows
  • flowoel() - Complete OEL implementation reference
  • Perfect for when you forget the sequence of steps!

✅ Artificial Neural Networks Module (114+ functions across 10 labs!)

• Lab 1: Tensor Operations - PyTorch basics, autograd, CUDA support
• Lab 2: Perceptron - sklearn implementation with decision boundaries
• Lab 3: ADALINE - Manual, semi-automatic, and PyTorch versions
• Lab 4: MLP - Multi-layer perceptron for classification and regression
• Lab 5: CNN - Convolutional neural networks for FashionMNIST
• Lab 6: CNN Filters - Custom filters with TensorFlow (edge, blur, sharpen)
• Lab 7: Transfer Learning - Pre-trained model fine-tuning
• Lab 8 (NEW!): Autoencoders - Undercomplete, Denoising, Convolutional
• Lab 9 (NEW!): RNN - Question-Answer, IMDB Sentiment Analysis
• Lab 10 (NEW!): LSTM - Next-word prediction, Text generation
• Lab Workflows: Complete code references with flowlab1() through flowlab10()

✅ Speech Processing Module (42 functions) - NEW in v0.1.6!

• Audio Loading: Load, play, save, resample audio files
• Spectrograms: Linear, Mel, Narrowband, Wideband spectrograms
• MFCC: Extract MFCC features (librosa + manual implementation)
• Vowel Synthesis: Generate vowel sounds with formant frequencies (OEL)
• Formant Analysis: Identify F1, F2, F3 formants automatically
• Pitch Analysis: Extract and visualize pitch histograms
• Dataset Utilities: Load datasets from ZIP/Drive with auto-preview
• Dataset Generation: Create synthetic vowel datasets
• Lab Workflows: Complete code for Labs 1, 2, 4, 6, OEL, Quiz
• OEL Functions: Perfect for teacher questions about vowel synthesis and analysis!

Requirements

Core (always installed)

• Python >= 3.7
• NumPy >= 1.21.0

Optional (install what you need)

# For Reinforcement Learning
pip install matplotlab[rl]

# For Artificial Neural Networks
pip install matplotlab[ann]

# For Speech Processing (NEW!)
pip install matplotlab[sp]

# Install everything
pip install matplotlab[all]

Individual module dependencies:

• RL: matplotlib, gymnasium, google-generativeai
• ANN: matplotlib, seaborn, pandas, torch, torchvision, tensorflow, scikit-learn, mlxtend, Pillow
• SP: matplotlib, librosa, soundfile, scipy, ipython, google-generativeai
• ANN: matplotlib, pytorch, tensorflow, scikit-learn
• SP: matplotlib, librosa, soundfile, scipy, ipython, google-generativeai

Key Design Philosophy

Simple, Beginner-Friendly Code:

• Uses nn.Sequential() for neural networks (no complex classes)
• Clear variable names: X_train, y_train, model, loss_fn
• Simple for loops and if-else statements
• No lambda functions or advanced Python features
• Easy to understand and modify

Documentation

• 171 total functions (44 RL + 85 ANN + 42 SP)
• Complete docstrings for every function
• Usage examples included
• All functions have .show() method to view source code
• See SP_MODULE_FUNCTIONS.md for complete SP function guide
• See DATASET_LOADING_GUIDE.md for dataset utilities

Quick Tips

# View source code of any function
sp.synthesize_vowel.show()
rl.policy_iteration.show()
ann.create_mlp_model.show()

# Get AI help
sp.query("How do I extract MFCC features?")
rl.query("Explain policy iteration")

License

MIT License - Free for educational use

Links

• PyPI: https://pypi.org/project/matplotlab/
• GitHub: https://github.com/Sohail-Creates/matplotlab/

---

For educational purposes | ML/RL implementations made simple