{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# Catastrophic forgetting on MNIST\n",
        "\n",
        "A companion to Lecture 14\n",
        "\n",
        "Marcel Turcotte  \n",
        "2026-10-27\n",
        "\n",
        "A network can learn a new task while losing the ability to solve an\n",
        "earlier one. This notebook trains a small MLP first on MNIST digits\n",
        "**0–4**, then on **5–9**. We measure both capabilities and compare\n",
        "training on new data alone with replay of a small memory of old training\n",
        "examples.\n",
        "\n",
        "This extends Lecture 14’s discussion of training challenges. The model\n",
        "has one fixed output layer with **ten logits** throughout; digit labels\n",
        "retain their original values. At evaluation, all ten classes compete.\n",
        "The experiment is a simple form of **class-incremental learning**, not\n",
        "an evaluation that tells the model which five classes to choose from.\n",
        "See <a href=\"https://arxiv.org/abs/1904.07734\" class=\"external\"\n",
        "data-icon=\"true\">van de Ven and Tolias (2019)</a>.\n",
        "\n",
        "## Data and a reproducible training budget\n",
        "\n",
        "We use MNIST’s official training/test separation. For a quick\n",
        "experiment, each digit contributes 1,000 training images and 200\n",
        "validation images from the official training partition. The official\n",
        "test images are not used for choosing settings. The first run downloads\n",
        "MNIST through Keras if it is not already cached.\n",
        "\n",
        "The seeds fix the initialization and sampling settings. Exact results\n",
        "can still vary with the backend and hardware; interpret the measured\n",
        "forgetting and retention in the executed run rather than expecting\n",
        "identical percentages on every system."
      ],
      "id": "ac5f029c-6b40-4172-a902-07e04c55c071"
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {},
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import keras\n",
        "\n",
        "SEED = 42\n",
        "EPOCHS = 6\n",
        "STEPS_PER_EPOCH = 40\n",
        "BATCH_SIZE = 128\n",
        "LEARNING_RATE = 0.1\n",
        "\n",
        "keras.utils.set_random_seed(SEED)\n",
        "(images, labels), (test_images, test_labels) = keras.datasets.mnist.load_data()\n",
        "data_rng = np.random.default_rng(SEED)\n",
        "train_indices, validation_indices = [], []\n",
        "for digit in range(10):\n",
        "    indices = data_rng.permutation(np.flatnonzero(labels == digit))\n",
        "    train_indices.extend(indices[:1000])\n",
        "    validation_indices.extend(indices[1000:1200])\n",
        "\n",
        "train_images, train_labels = images[train_indices], labels[train_indices]\n",
        "validation_images = images[validation_indices]\n",
        "validation_labels = labels[validation_indices]\n",
        "train_a = (train_images[train_labels < 5], train_labels[train_labels < 5])\n",
        "train_b = (train_images[train_labels >= 5], train_labels[train_labels >= 5])\n",
        "validation_a = (validation_images[validation_labels < 5], validation_labels[validation_labels < 5])\n",
        "validation_b = (validation_images[validation_labels >= 5], validation_labels[validation_labels >= 5])"
      ],
      "id": "4217d50c"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## One model, ten possible classes\n",
        "\n",
        "The hidden layer uses ReLU and the output is linear.\n",
        "`SparseCategoricalCrossentropy(from_logits=True)` receives the logits\n",
        "directly and accepts integer digit labels. Softmax would convert those\n",
        "logits to probabilities; `argmax` gives the same predicted class before\n",
        "or after softmax.\n",
        "\n",
        "SGD uses a constant learning rate with no momentum or regularization.\n",
        "This keeps the comparison focused on the training data supplied during\n",
        "the second phase."
      ],
      "id": "b6a9b1cb-5666-47e4-8871-166885b5db08"
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {},
      "outputs": [],
      "source": [
        "def compile_model(model):\n",
        "    model.compile(optimizer=keras.optimizers.SGD(learning_rate=LEARNING_RATE),\n",
        "                  loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True))\n",
        "\n",
        "def make_model():\n",
        "    model = keras.Sequential([\n",
        "        keras.layers.Input(shape=(28, 28)),\n",
        "        keras.layers.Rescaling(1/255),\n",
        "        keras.layers.Flatten(),\n",
        "        keras.layers.Dense(128, activation='relu', kernel_initializer='he_normal'),\n",
        "        keras.layers.Dense(10)\n",
        "    ])\n",
        "    compile_model(model)\n",
        "    return model\n",
        "\n",
        "def predicted_classes(model, x):\n",
        "    logits = keras.ops.convert_to_numpy(model(x, training=False))\n",
        "    return logits.argmax(axis=1)\n",
        "\n",
        "def validation_accuracy(model):\n",
        "    return [np.mean(predicted_classes(model, x) == y)\n",
        "            for x, y in (validation_a, validation_b)]"
      ],
      "id": "ee75a921"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## The same training loop, with or without memory\n",
        "\n",
        "Every phase has the same number of parameter updates and the same batch\n",
        "size. Sampling with replacement makes that update budget explicit.\n",
        "Without memory, each batch contains new-task examples only. With memory,\n",
        "half of each batch comes from the new task and half from the stored old\n",
        "examples. Replay therefore changes the data mixture within the fixed\n",
        "budget.\n",
        "\n",
        "The returned history records validation accuracy for both groups before\n",
        "training and after each epoch. `train_on_batch` performs the forward\n",
        "pass, gradient computation, and SGD update; the NumPy code only\n",
        "constructs the batches."
      ],
      "id": "262ef3d3-9247-4f02-b76b-225a51806733"
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {},
      "outputs": [],
      "source": [
        "def train_phase(model, training_data, seed, memory=None):\n",
        "    rng = np.random.default_rng(seed)\n",
        "    x, y = training_data\n",
        "    history = [validation_accuracy(model)]\n",
        "    new_count = BATCH_SIZE if memory is None else BATCH_SIZE // 2\n",
        "    for epoch in range(EPOCHS):\n",
        "        for _ in range(STEPS_PER_EPOCH):\n",
        "            indices = rng.integers(len(y), size=new_count)\n",
        "            xb, yb = x[indices], y[indices]\n",
        "            if memory is not None:\n",
        "                old_x, old_y = memory\n",
        "                old_indices = rng.integers(len(old_y), size=BATCH_SIZE-new_count)\n",
        "                xb = np.concatenate([xb, old_x[old_indices]])\n",
        "                yb = np.concatenate([yb, old_y[old_indices]])\n",
        "            model.train_on_batch(xb, yb)\n",
        "        history.append(validation_accuracy(model))\n",
        "    return np.asarray(history)"
      ],
      "id": "f96e22a4"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Phase A: learn digits 0–4"
      ],
      "id": "63994284-cf4a-47d8-ab93-1cb40febbeb0"
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {},
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Validation accuracy after A:\n",
            "  digits 0–4: 95.6%\n",
            "  digits 5–9: 0.0%"
          ]
        }
      ],
      "source": [
        "sequential_model = make_model()\n",
        "history_a = train_phase(sequential_model, train_a, seed=101)\n",
        "weights_after_a = [weight.copy() for weight in sequential_model.get_weights()]\n",
        "test_predictions_after_a = predicted_classes(sequential_model, test_images)\n",
        "print('Validation accuracy after A:')\n",
        "print(f'  digits 0–4: {history_a[-1, 0]:.1%}')\n",
        "print(f'  digits 5–9: {history_a[-1, 1]:.1%}')"
      ],
      "id": "b760c3a5"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Poor performance on digits 5–9 at this point is expected: they have not\n",
        "been training targets. The reference for forgetting is the model’s\n",
        "performance on digits 0–4 **after learning A**.\n",
        "\n",
        "## Phase B: new data only versus replay\n",
        "\n",
        "The sequential branch continues from its learned parameters. The replay\n",
        "branch starts from an identical copy of those parameters. Both branches\n",
        "use a fresh SGD optimizer with the same constant learning rate;\n",
        "recompiling the sequential branch does not reinitialize its model\n",
        "weights.\n",
        "\n",
        "The memory stores twenty training images per old class, for one hundred\n",
        "images in total. No validation or test image is placed in memory. The\n",
        "same stage-B update count is used in both branches."
      ],
      "id": "2c64a251-068e-41a0-b7f9-b3f217bac9b3"
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {},
      "outputs": [],
      "source": [
        "memory_rng = np.random.default_rng(202)\n",
        "memory_indices = np.concatenate([\n",
        "    memory_rng.choice(np.flatnonzero(train_a[1] == digit), size=20, replace=False)\n",
        "    for digit in range(5)\n",
        "])\n",
        "memory = (train_a[0][memory_indices], train_a[1][memory_indices])\n",
        "\n",
        "replay_model = make_model()\n",
        "replay_model.set_weights(weights_after_a)\n",
        "compile_model(sequential_model)\n",
        "history_b = train_phase(sequential_model, train_b, seed=303)\n",
        "history_replay = train_phase(replay_model, train_b, seed=303, memory=memory)"
      ],
      "id": "6786778a"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Follow both capabilities"
      ],
      "id": "4cea7775-5560-42eb-9f0f-dd7c00952400"
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {},
      "outputs": [
        {
          "output_type": "display_data",
          "metadata": {},
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABBYAAAFZCAYAAAAsFKy0AAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90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\n"
          }
        }
      ],
      "source": [
        "fig, axes = plt.subplots(1, 2, figsize=(11, 3.7), sharey=True)\n",
        "for ax, continuation, title in zip(axes, (history_b, history_replay),\n",
        "                                  ('B only', 'B plus replay of 100 A images')):\n",
        "    history = np.vstack([history_a, continuation[1:]])\n",
        "    epochs = np.arange(len(history))\n",
        "    ax.plot(epochs, 100*history[:, 0], 'o-', label='Digits 0–4 (A)')\n",
        "    ax.plot(epochs, 100*history[:, 1], 'o-', label='Digits 5–9 (B)')\n",
        "    ax.axvline(EPOCHS, color='black', linestyle='--', label='Start phase B')\n",
        "    ax.set(title=title, xlabel='Epoch', ylim=(0, 100), xticks=np.arange(0, 2*EPOCHS+1, 2))\n",
        "    ax.grid(alpha=0.25)\n",
        "    ax.legend(fontsize=8)\n",
        "axes[0].set_ylabel('Validation accuracy (%)')\n",
        "fig.tight_layout()\n",
        "fig.savefig('catastrophic_forgetting_accuracy.png', dpi=150, bbox_inches='tight')\n",
        "plt.show()"
      ],
      "id": "9914b0ba"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "The two panels share the same phase-A history. Read the second phase in\n",
        "two directions: does the model learn B, and how much of A does it\n",
        "retain? The validation curves describe this fixed experiment; they are\n",
        "not used here to select hyperparameters or stopping times.\n",
        "\n",
        "Replay reduces forgetting in this experiment but does not eliminate it.\n",
        "Some of the fixed phase-B update budget is devoted to A, so both\n",
        "retention and new-task performance should be considered when comparing\n",
        "the branches.\n",
        "\n",
        "## Stage-end measurements on the official test set\n",
        "\n",
        "Report the groups separately: aggregate accuracy can hide a large loss\n",
        "of an earlier capability. All ten output classes remain available for\n",
        "every prediction."
      ],
      "id": "688646d4-6596-446c-81dc-726c98225613"
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {},
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Checkpoint                A (0–4)    B (5–9)   All digits\n",
            "After A                     97.1%       0.0%      49.9%\n",
            "After B only                 2.3%      93.9%      46.8%\n",
            "After B with replay         79.8%      91.4%      85.4%\n",
            "A accuracy drop, After B only: 94.8 percentage points\n",
            "A accuracy drop, After B with replay: 17.3 percentage points"
          ]
        }
      ],
      "source": [
        "test_predictions_sequential = predicted_classes(sequential_model, test_images)\n",
        "test_predictions_replay = predicted_classes(replay_model, test_images)\n",
        "mask_a = test_labels < 5\n",
        "\n",
        "def group_accuracies(predictions):\n",
        "    correct = predictions == test_labels\n",
        "    return np.array([correct[mask_a].mean(), correct[~mask_a].mean(), correct.mean()])\n",
        "\n",
        "test_results = {\n",
        "    'After A': group_accuracies(test_predictions_after_a),\n",
        "    'After B only': group_accuracies(test_predictions_sequential),\n",
        "    'After B with replay': group_accuracies(test_predictions_replay)\n",
        "}\n",
        "print(f\"{'Checkpoint':<22} {'A (0–4)':>10} {'B (5–9)':>10} {'All digits':>12}\")\n",
        "for name, values in test_results.items():\n",
        "    print(f'{name:<22}' + ''.join(f'{100*value:>10.1f}%' for value in values))\n",
        "for name in ('After B only', 'After B with replay'):\n",
        "    forgetting_pp = 100*(test_results['After A'][0]-test_results[name][0])\n",
        "    print(f'A accuracy drop, {name}: {forgetting_pp:.1f} percentage points')"
      ],
      "id": "4b463930"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "These measurements compare predeclared checkpoints. The phase-A test\n",
        "predictions were recorded before phase-B training and are not recomputed\n",
        "using later parameters.\n",
        "\n",
        "## The same images, different predictions"
      ],
      "id": "148a2901-74b6-4cc0-af9e-778ef2671ba1"
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {},
      "outputs": [
        {
          "output_type": "display_data",
          "metadata": {},
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABHYAAAD3CAYAAAB4ktPuAAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90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\nAAAAAAAAEUVgBwAAAAAAwImm/we5NYijiaX0pgAAAABJRU5ErkJggg==\n"
          }
        }
      ],
      "source": [
        "forgotten = np.flatnonzero(mask_a & (test_predictions_after_a == test_labels)\n",
        "                           & (test_predictions_sequential != test_labels))[:6]\n",
        "if len(forgotten):\n",
        "    fig, axes = plt.subplots(1, len(forgotten), figsize=(2*len(forgotten), 2.6), squeeze=False)\n",
        "    for ax, index in zip(axes[0], forgotten):\n",
        "        ax.imshow(test_images[index], cmap='gray')\n",
        "        ax.set_title(f'True: {test_labels[index]}\\nAfter A: {test_predictions_after_a[index]}\\nAfter B: {test_predictions_sequential[index]}', fontsize=10)\n",
        "        ax.axis('off')\n",
        "    fig.tight_layout()\n",
        "    plt.show()\n",
        "else:\n",
        "    print('No previously correct A test images became incorrect in this run.')"
      ],
      "id": "0bd7d5df"
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "The inputs and their labels did not change. The parameters changed\n",
        "during further training. In this disjoint-class example, the decline can\n",
        "include a strong bias toward recent classes as well as changes in shared\n",
        "features; it does not establish that every useful old representation has\n",
        "been erased.\n",
        "\n",
        "## What does this imply for deployed models?\n",
        "\n",
        "Prediction with a trained model normally leaves its weights fixed.\n",
        "Learning requires an explicit update process. Catastrophic forgetting\n",
        "can occur during online learning or during periodic batch fine-tuning\n",
        "when new updates do not protect earlier behavior.\n",
        "\n",
        "Production neural networks can learn online;\n",
        "<a href=\"https://arxiv.org/abs/2209.07663\" class=\"external\"\n",
        "data-icon=\"true\">Monolith</a> is a real-time recommendation example. The\n",
        "lesson is to design and evaluate updates carefully, not that online\n",
        "learning is impossible. <a\n",
        "href=\"https://developers.google.com/machine-learning/crash-course/production-ml-systems/static-vs-dynamic-training\"\n",
        "class=\"external\" data-icon=\"true\">Google’s production-ML discussion</a>\n",
        "distinguishes static and dynamic training.\n",
        "\n",
        "Replay explicitly reintroduces an objective on earlier data. Ordinary\n",
        "zero-centered L2 penalizes large weights; it does not explicitly\n",
        "preserve previously learned predictions. Methods such as\n",
        "<a href=\"https://arxiv.org/abs/1612.00796\" class=\"external\"\n",
        "data-icon=\"true\">elastic weight consolidation</a> use other mechanisms\n",
        "to protect earlier learning. No continual-learning method is implemented\n",
        "or claimed to be universally best here.\n",
        "\n",
        "For reflection: why can B accuracy rise while A accuracy falls? Why is\n",
        "reporting only the mean accuracy insufficient? What changes when old\n",
        "examples occupy part of the fixed stage-B update budget?\n",
        "\n",
        "[Return to Lecture 14](index.qmd)."
      ],
      "id": "18791092-6329-49be-896f-69157a607351"
    }
  ],
  "nbformat": 4,
  "nbformat_minor": 5,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3 (ipykernel)",
      "language": "python",
      "path": "/Users/turcotte/.virtualenvs/csi4106/share/jupyter/kernels/python3"
    },
    "language_info": {
      "name": "python",
      "codemirror_mode": {
        "name": "ipython",
        "version": "3"
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.12.12"
    }
  }
}