{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# Embedding Visualization\n",
        "\n",
        "## Overview\n",
        "\n",
        "This notebook demonstrates how to generate embeddings, optimize them, and visualize using t-SNE, PCA, and UMAP dimensionality reduction techniques.\n",
        "\n",
        "### Learning Objectives\n",
        "\n",
        "- Generate embeddings for documents\n",
        "- Optimize embeddings for better quality\n",
        "- Visualize embeddings using t-SNE\n",
        "- Visualize embeddings using PCA\n",
        "- Visualize embeddings using UMAP\n",
        "\n",
        "---\n",
        "\n",
        "## Workflow\n",
        "\n",
        "**Generate Embeddings → Optimize → Visualize**\n",
        "\n",
        "Visualization helps understand the structure and relationships in embedding spaces.\n",
        "\n",
        "---\n",
        "\n",
        "## Step 1: Generate Embeddings\n",
        "\n",
        "Start by generating embeddings for your documents.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "from semantica.embeddings import EmbeddingGenerator\n",
        "import numpy as np\n",
        "\n",
        "documents = [\n",
        "    \"Machine learning algorithms\",\n",
        "    \"Deep neural networks\",\n",
        "    \"Natural language processing\",\n",
        "    \"Computer vision\",\n",
        "    \"Reinforcement learning\",\n",
        "]\n",
        "\n",
        "generator = EmbeddingGenerator()\n",
        "\n",
        "try:\n",
        "    embeddings = generator.generate(documents)\n",
        "    print(\"✓ Embeddings generated\")\n",
        "    print(f\"  Documents: {len(documents)}\")\n",
        "    print(f\"  Embedding dimension: {embeddings.shape[1] if hasattr(embeddings, 'shape') else 'N/A'}\")\n",
        "    \n",
        "except Exception as e:\n",
        "    print(f\"✗ Error generating embeddings: {e}\")\n",
        "    embeddings = np.random.rand(len(documents), 1536).astype(np.float32)\n",
        "    print(\"  Using demo embeddings\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Step 2: Optimize Embeddings\n",
        "\n",
        "Optimize embeddings to improve their quality and reduce noise.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "from semantica.embeddings import EmbeddingOptimizer\n",
        "\n",
        "optimizer = EmbeddingOptimizer()\n",
        "\n",
        "try:\n",
        "    optimized_embeddings = optimizer.optimize(embeddings)\n",
        "    print(\"✓ Embeddings optimized\")\n",
        "    print(f\"  Optimized embeddings ready for visualization\")\n",
        "    \n",
        "except Exception as e:\n",
        "    print(f\"✗ Error optimizing embeddings: {e}\")\n",
        "    optimized_embeddings = embeddings\n",
        "    print(\"  Using original embeddings\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Step 3: Visualize with t-SNE\n",
        "\n",
        "Use t-SNE (t-Distributed Stochastic Neighbor Embedding) to visualize embeddings in 2D space.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "from semantica.visualization import EmbeddingVisualizer\n",
        "\n",
        "visualizer = EmbeddingVisualizer()\n",
        "\n",
        "labels = [f\"Doc {i+1}\" for i in range(len(documents))]\n",
        "\n",
        "try:\n",
        "    visualizer.visualize_tsne(optimized_embeddings, labels)\n",
        "    print(\"✓ t-SNE visualization complete\")\n",
        "    print(\"  Note: t-SNE shows local structure and clusters in embedding space\")\n",
        "    \n",
        "except Exception as e:\n",
        "    print(f\"✗ Error visualizing with t-SNE: {e}\")\n",
        "    print(\"  Note: t-SNE reduces high-dimensional embeddings to 2D for visualization\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Step 4: Visualize with PCA\n",
        "\n",
        "Use PCA (Principal Component Analysis) to visualize embeddings, preserving global structure.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "try:\n",
        "    visualizer.visualize_pca(optimized_embeddings, labels)\n",
        "    print(\"✓ PCA visualization complete\")\n",
        "    print(\"  Note: PCA preserves global structure and variance\")\n",
        "    \n",
        "except Exception as e:\n",
        "    print(f\"✗ Error visualizing with PCA: {e}\")\n",
        "    print(\"  Note: PCA reduces dimensions while preserving maximum variance\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Step 5: Visualize with UMAP\n",
        "\n",
        "Use UMAP (Uniform Manifold Approximation and Projection) for a balance between local and global structure.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "try:\n",
        "    visualizer.visualize_umap(optimized_embeddings, labels)\n",
        "    print(\"✓ UMAP visualization complete\")\n",
        "    print(\"  Note: UMAP balances local and global structure preservation\")\n",
        "    print(\"\\n✓ Embedding visualization complete\")\n",
        "    print(\"  All visualization methods demonstrate different aspects of embedding space\")\n",
        "    \n",
        "except Exception as e:\n",
        "    print(f\"✗ Error visualizing with UMAP: {e}\")\n",
        "    print(\"  Note: UMAP provides a good balance between t-SNE and PCA\")\n"
      ]
    }
  ],
  "metadata": {
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}