neuromatch · iamzoltan · Jul 18, 2026 · Jul 2, 2026 · Jul 2, 2026 · Jul 2, 2026
diff --git a/projects/project-notebooks/Microlearning.ipynb b/projects/project-notebooks/Microlearning.ipynb
diff --git a/requirements.txt b/requirements.txt
@@ -33,3 +33,41 @@ git+https://github.com/neuromatch/pyBPL#egg=pybpl
 git+https://github.com/neuromatch/MotorNet#egg=motornet
 git+https://github.com/ctn-waterloo/sspspace@neuromatch#egg=sspspace
 git+https://github.com/mitchellostrow/DSA#egg=dsa-metric
+
+datatops
+
+dm_acme
+
+dm_control
+
+graphviz
+
+ipympl
+
+ipysankeywidget
+
+ipython
+
+memory_profiler
+
+mpl_interactions
+
+nengo
+
+nengo_gui
+
+nengo_spa
+
+pandas
+
+pickleshare
+
+tabulate
+
+torch_optimizer
+
+torchmetrics
+
+trdg
+
+xarray
diff --git a/tutorials/Bonus_BonusContent/Bonus_Tutorial2.ipynb b/tutorials/Bonus_BonusContent/Bonus_Tutorial2.ipynb
@@ -2,13 +2,14 @@
  "cells": [
   {
    "cell_type": "markdown",
+   "id": "b00e9b5d",
    "metadata": {
     "colab_type": "text",
     "execution": {},
     "id": "view-in-github"
    },
    "source": [
-    "<a href=\"https://colab.research.google.com/github/NeuromatchAcademy/course-content/blob/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/NeuromatchAcademy/course-content/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
+    "<a href=\"https://colab.research.google.com/github/neuromatch/NeuroAI_Course/blob/main/tutorials/Bonus_BonusContent/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/neuromatch/NeuroAI_Course/main/tutorials/Bonus_BonusContent/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
    ]
   },
   {
@@ -79,7 +80,10 @@
    "outputs": [],
    "source": [
     "# @title Install dependecies\n",
-    "!pip install plotly --quiet"
+    "\n",
+    "from os import environ as env\n",
+    "if any(var in env for var in [\"COLAB_RELEASE_TAG\", \"KAGGLE_KERNEL_RUN_TYPE\"]):\n",
+    "    %pip install plotly --quiet"
    ]
   },
   {
@@ -957,15 +961,7 @@
     "\n",
     "The shallow autoencoder has a compression ratio of **784:2 = 392:1**. The first additional hidden layer has a compression ratio of **2:1**,  followed by a hidden layer that sets the bottleneck compression ratio of **32:1**.\n",
     "\n",
-    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
+    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
     "\n",
     "```python\n",
     "model[:-2].apply(init_weights_kaiming_normal)\n",
@@ -1125,8 +1121,7 @@
     "The encoder map shows well-separated clusters that correspond to the associated digits in the decoder grid. The decoder grid also shows that the network is robust to digit skewness, i.e., digits leaning to the left or the right are recognized in the same digit class.\n",
     "\n",
     "**Instructions:**\n",
-    "* Please execute the cells below\n",
-    "\n"
+    "* Please execute the cells below"
    ]
   },
   {
@@ -1730,7 +1725,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.18"
+   "version": "3.10.20"
   }
  },
  "nbformat": 4,

diff --git a/tutorials/Bonus_BonusContent/instructor/Bonus_Tutorial2.ipynb b/tutorials/Bonus_BonusContent/instructor/Bonus_Tutorial2.ipynb
@@ -2,13 +2,14 @@
  "cells": [
   {
    "cell_type": "markdown",
+   "id": "dee4365d",
    "metadata": {
     "colab_type": "text",
     "execution": {},
     "id": "view-in-github"
    },
    "source": [
-    "<a href=\"https://colab.research.google.com/github/NeuromatchAcademy/course-content/blob/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/NeuromatchAcademy/course-content/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
+    "<a href=\"https://colab.research.google.com/github/neuromatch/NeuroAI_Course/blob/main/tutorials/Bonus_BonusContent/instructor/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/neuromatch/NeuroAI_Course/main/tutorials/Bonus_BonusContent/instructor/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
    ]
   },
   {
@@ -79,7 +80,10 @@
    "outputs": [],
    "source": [
     "# @title Install dependecies\n",
-    "!pip install plotly --quiet"
+    "\n",
+    "from os import environ as env\n",
+    "if any(var in env for var in [\"COLAB_RELEASE_TAG\", \"KAGGLE_KERNEL_RUN_TYPE\"]):\n",
+    "    %pip install plotly --quiet"
    ]
   },
   {
@@ -957,15 +961,7 @@
     "\n",
     "The shallow autoencoder has a compression ratio of **784:2 = 392:1**. The first additional hidden layer has a compression ratio of **2:1**,  followed by a hidden layer that sets the bottleneck compression ratio of **32:1**.\n",
     "\n",
-    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
+    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
     "\n",
     "```python\n",
     "model[:-2].apply(init_weights_kaiming_normal)\n",
@@ -1127,8 +1123,7 @@
     "The encoder map shows well-separated clusters that correspond to the associated digits in the decoder grid. The decoder grid also shows that the network is robust to digit skewness, i.e., digits leaning to the left or the right are recognized in the same digit class.\n",
     "\n",
     "**Instructions:**\n",
-    "* Please execute the cells below\n",
-    "\n"
+    "* Please execute the cells below"
    ]
   },
   {
@@ -1734,7 +1729,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.18"
+   "version": "3.10.20"
   }
  },
  "nbformat": 4,

diff --git a/tutorials/Bonus_BonusContent/student/Bonus_Tutorial2.ipynb b/tutorials/Bonus_BonusContent/student/Bonus_Tutorial2.ipynb
@@ -2,13 +2,14 @@
  "cells": [
   {
    "cell_type": "markdown",
+   "id": "eab90967",
    "metadata": {
     "colab_type": "text",
     "execution": {},
     "id": "view-in-github"
    },
    "source": [
-    "<a href=\"https://colab.research.google.com/github/NeuromatchAcademy/course-content/blob/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/NeuromatchAcademy/course-content/main/tutorials/Bonus_Autoencoders/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
+    "<a href=\"https://colab.research.google.com/github/neuromatch/NeuroAI_Course/blob/main/tutorials/Bonus_BonusContent/student/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/neuromatch/NeuroAI_Course/main/tutorials/Bonus_BonusContent/student/Bonus_Tutorial2.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
    ]
   },
   {
@@ -79,7 +80,10 @@
    "outputs": [],
    "source": [
     "# @title Install dependecies\n",
-    "!pip install plotly --quiet"
+    "\n",
+    "from os import environ as env\n",
+    "if any(var in env for var in [\"COLAB_RELEASE_TAG\", \"KAGGLE_KERNEL_RUN_TYPE\"]):\n",
+    "    %pip install plotly --quiet"
    ]
   },
   {
@@ -957,15 +961,7 @@
     "\n",
     "The shallow autoencoder has a compression ratio of **784:2 = 392:1**. The first additional hidden layer has a compression ratio of **2:1**,  followed by a hidden layer that sets the bottleneck compression ratio of **32:1**.\n",
     "\n",
-    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.\n",
-    "\n",
-    "&nbsp;\n",
-    "\n",
-    "Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
+    "The choice of hidden layer size aims to reduce the compression rate in the bottleneck layer while increasing the count of trainable parameters.  For example, if the compression rate of the first hidden layer doubles from **2:1** to **4:1**, the count of trainable parameters halves from 667K to 333K.This deep autoencoder's performance may be further improved by adding additional hidden layers and by increasing the count of trainable parameters in each layer. These improvements have a diminishing return due to challenges associated with training under high parameter count and depth. One option explored in the *Bonus* section is to add a first hidden layer with 2x - 3x the input size. This size increase results in millions of parameters at the cost of longer training time.Weight initialization is particularly important in deep networks. The availability of large datasets and weight initialization likely drove the deep learning revolution of 2010. We'll implement Kaiming normal as follows:\n",
     "\n",
     "```python\n",
     "model[:-2].apply(init_weights_kaiming_normal)\n",
@@ -1035,8 +1031,7 @@
     "execution": {}
    },
    "source": [
-    "[*Click for solution*](https://github.com/neuromatch/course-content-template/tree/main/tutorials/Bonus_BonusContent/solutions/Bonus_Tutorial2_Solution_0d50004f.py)\n",
-    "\n"
+    "[*Click for solution*](https://github.com/neuromatch/NeuroAI_Course/tree/main/tutorials/Bonus_BonusContent/solutions/Bonus_Tutorial2_Solution_0d50004f.py)"
    ]
   },
   {
@@ -1087,8 +1082,7 @@
     "The encoder map shows well-separated clusters that correspond to the associated digits in the decoder grid. The decoder grid also shows that the network is robust to digit skewness, i.e., digits leaning to the left or the right are recognized in the same digit class.\n",
     "\n",
     "**Instructions:**\n",
-    "* Please execute the cells below\n",
-    "\n"
+    "* Please execute the cells below"
    ]
   },
   {
@@ -1356,8 +1350,7 @@
     "execution": {}
    },
    "source": [
-    "[*Click for solution*](https://github.com/neuromatch/course-content-template/tree/main/tutorials/Bonus_BonusContent/solutions/Bonus_Tutorial2_Solution_f82e3b9c.py)\n",
-    "\n"
+    "[*Click for solution*](https://github.com/neuromatch/NeuroAI_Course/tree/main/tutorials/Bonus_BonusContent/solutions/Bonus_Tutorial2_Solution_f82e3b9c.py)"
    ]
   },
   {
@@ -1661,7 +1654,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.18"
+   "version": "3.10.20"
   }
  },
  "nbformat": 4,

diff --git a/tutorials/W1D1_Generalization/W1D1_Intro.ipynb b/tutorials/W1D1_Generalization/W1D1_Intro.ipynb
@@ -2,12 +2,14 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "id": "4e3d2b26-a059-4683-8bfd-2499a50eb346",
+   "id": "7dba9cef",
    "metadata": {
-    "execution": {}
+    "colab_type": "text",
+    "execution": {},
+    "id": "view-in-github"
    },
    "source": [
-    "<a href=\"https://colab.research.google.com/github/neuromatch/NeuroAI_Course/blob/main/tutorials/W1D1_Generalization/student/W1D1_Intro.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/neuromatch/NeuroAI_Course/main/tutorials/W1D1_Generalization/student/W1D1_Intro.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
+    "<a href=\"https://colab.research.google.com/github/neuromatch/NeuroAI_Course/blob/main/tutorials/W1D1_Generalization/W1D1_Intro.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> &nbsp; <a href=\"https://kaggle.com/kernels/welcome?src=https://raw.githubusercontent.com/neuromatch/NeuroAI_Course/main/tutorials/W1D1_Generalization/W1D1_Intro.ipynb\" target=\"_parent\"><img src=\"https://kaggle.com/static/images/open-in-kaggle.svg\" alt=\"Open in Kaggle\"/></a>"
    ]
   },
   {
@@ -33,7 +35,9 @@
    "source": [
     "# @title Install and import feedback gadget\n",
     "\n",
-    "!pip install vibecheck datatops --quiet\n",
+    "from os import environ as env\n",
+    "if any(var in env for var in [\"COLAB_RELEASE_TAG\", \"KAGGLE_KERNEL_RUN_TYPE\"]):\n",
+    "    %pip install vibecheck datatops --quiet\n",
     "\n",
     "from vibecheck import DatatopsContentReviewContainer\n",
     "def content_review(notebook_section: str):\n",
@@ -232,7 +236,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.19"
+   "version": "3.10.20"
   }
  },
  "nbformat": 4,