Merge pull request #417 from neuromatch/W1D3-superset

W1D3 Post-Course Update (Superset)
neuromatch · Aug 21, 2024 · 5ad4631 · 5ad4631
2 parents a7fea07 + f5b9eda
commit 5ad4631
Show file tree

Hide file tree

Showing 17 changed files with 75 additions and 206 deletions.
diff --git a/requirements.txt b/requirements.txt
@@ -23,7 +23,7 @@ tensorflow==2.8
 ffmpeg
 imageio-ffmpeg
 torchlens
-rsatoolbox
+rsatoolbox==0.1.5
 vibecheck
 seaborn
 plotly

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial1.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial1.ipynb
@@ -113,8 +113,9 @@
    "source": [
     "# @title Install dependencies\n",
     "\n",
-    "!pip install -q ipympl ipywidgets mpl_interactions[\"jupyter\"] rsatoolbox torchlens\n",
+    "!pip install -q ipympl ipywidgets mpl_interactions[\"jupyter\"] torchlens\n",
     "!pip install -q graphviz\n",
+    "!pip install rsatoolbox==0.1.5\n",
     "\n",
     "# To install jupyter-matplotlib (ipympl) via pip\n",
     "!pip install -q torchlens\n",

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial2.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial2.ipynb
@@ -90,7 +90,8 @@
    "source": [
     "# @title Install and import feedback gadget\n",
     "\n",
-    "!pip install torch torchvision matplotlib numpy scikit-learn rsatoolbox scipy vibecheck --quiet\n",
+    "!pip install torch torchvision matplotlib numpy scikit-learn scipy vibecheck --quiet\n",
+    "!pip install rsatoolbox==0.1.5 --quiet\n",
     "\n",
     "from vibecheck import DatatopsContentReviewContainer\n",
     "def content_review(notebook_section: str):\n",
@@ -1269,7 +1270,7 @@
    "source": [
     "# @title Representational Path\n",
     "\n",
-    "imgs, labels = next(iter(test_loader)) #grab 500 samples from the test set\n",
+    "imgs, labels = next(iter(test_loader))\n",
     "\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "\n",
@@ -1330,7 +1331,6 @@
    "source": [
     "# @title Test your idea by executing this cell!\n",
     "\n",
-    "#imgs, labels = next(iter(test_loader)) #grab 500 samples from the test set\n",
     "imgs, labels = sample_images(test_loader, n=5)\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "\n",
@@ -1476,43 +1476,24 @@
     "execution": {}
    },
    "source": [
-    "## Coding Exercise 1: Dimensionality reduction visualization\n",
+    "## Dimensionality reduction visualization\n",
     "\n",
-    "Use MDS to visualize the changes in the representational geometry across the layers of this network in response to original MNIST test images."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "execution": {}
-   },
-   "outputs": [],
-   "source": [
-    "#################################################\n",
-    "# Fill out and remove\n",
-    "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
-    "#################################################\n",
-    "\n",
-    "return_layers = ['input', 'conv1', 'conv2', 'fc1', 'fc2']\n",
-    "imgs, labels = sample_images(test_loader, n=50) # grab 500 samples from the test set\n",
-    "model_features = ...\n",
-    "\n",
-    "plot_dim_reduction(model_features, labels, transformer_funcs =['MDS'])"
+    "Now, we are going to use MDS to visualize the changes in the representational geometry across the layers of this network in response to the original MNIST test images."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
+    "cellView": "form",
     "execution": {}
    },
    "outputs": [],
    "source": [
-    "# to_remove solution\n",
+    "# @title Plot the representational geometry changes\n",
     "\n",
     "return_layers = ['input', 'conv1', 'conv2', 'fc1', 'fc2']\n",
-    "imgs, labels = sample_images(test_loader, n=50) # grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) # grab 50 samples from the test set\n",
     "model_features = extract_features(model_robust, imgs.to(device), return_layers)\n",
     "\n",
     "plot_dim_reduction(model_features, labels, transformer_funcs =['MDS'])"
@@ -1556,7 +1537,7 @@
    },
    "outputs": [],
    "source": [
-    "imgs, labels = sample_images(test_loader, n=50) #grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) #grab 50 samples from the test set\n",
     "\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "model_features_robust = extract_features(model_robust, imgs.to(device), return_layers='all')\n",
@@ -1583,7 +1564,7 @@
    },
    "outputs": [],
    "source": [
-    "imgs, labels = sample_images(test_loader, n=50) #grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) #grab 50 samples from the test set\n",
     "\n",
     "adv_imgs = generate_adversarial(model, imgs.to(device), labels.to(device), eps)\n",
     "\n",
@@ -1623,33 +1604,9 @@
     "execution": {}
    },
    "source": [
-    "## Coding Exercise 2: Compare representational paths\n",
-    "\n",
-    "Train another instance of the model on the original MNIST data for only one epoch with a different seed. Then compare the representational paths of the two instances that are both trained on the original MNIST with one another. Display the paths in blue and cyan colors."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "execution": {}
-   },
-   "outputs": [],
-   "source": [
-    "#################################################\n",
-    "## TODO for students: fill in the missing variables ##\n",
-    "# Fill out function and remove\n",
-    "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
-    "#################################################\n",
+    "## Compare representational paths\n",
     "\n",
-    "args = build_args()\n",
-    "torch.manual_seed(args.seed+1)\n",
-    "args.epochs = 1\n",
-    "#build_model\n",
-    "model_new_seed = Net().to(device)\n",
-    "\n",
-    "optimizer = optim.Adadelta(...)\n",
-    "train_model(args, model_new_seed, optimizer)"
+    "Train another instance of the model on the original MNIST data for only one epoch with a different seed (the basic model is trained for one epoch too). Then compare the representational paths of the two instances that are both trained on the original MNIST with one another. Display the paths in blue and cyan colors."
    ]
   },
   {
@@ -1660,8 +1617,6 @@
    },
    "outputs": [],
    "source": [
-    "# to_remove solution\n",
-    "\n",
     "args = build_args()\n",
     "torch.manual_seed(args.seed+1)\n",
     "args.epochs = 1\n",

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial3.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial3.ipynb
@@ -102,7 +102,9 @@
    "source": [
     "# @title Install and import feedback gadget\n",
     "\n",
-    "!pip install numpy pandas torch torchvision matplotlib ipython Pillow rsatoolbox plotly networkx requests vibecheck --quiet\n",
+    "!pip install numpy pandas torch torchvision matplotlib ipython Pillow plotly networkx requests vibecheck --quiet\n",
+    "!pip install rsatoolbox==0.1.5 --quiet\n",
+    "\n",
     "\n",
     "from vibecheck import DatatopsContentReviewContainer\n",
     "def content_review(notebook_section: str):\n",
@@ -1676,7 +1678,7 @@
    "source": [
     "## Coding Exercise 1: RDMs of AlexNet\n",
     "\n",
-    "Use the RSA toolbox to compute the RDMs for the layers of AlexNet."
+    "Use the RSA toolbox to compute the RDMs for the layers of AlexNet. It should be done in the very same way as RDMs for fMRI patterns above."
    ]
   },
   {
@@ -1820,7 +1822,7 @@
     "# Fill out function and remove\n",
     "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
     "#################################################\n",
-    "v1_rdms = ...\n",
+    "v1_rdms = fmri_rdms.subset('roi', ...)\n",
     "show_rdm_plotly(v1_rdms, rdm_descriptor='subject')"
    ]
   },

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial4.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/W1D3_Tutorial4.ipynb
@@ -112,7 +112,8 @@
    "source": [
     "# @title Install dependencies and import feedback gadget\n",
     "\n",
-    "!pip install numpy xarray scipy scikit-learn matplotlib seaborn tqdm rsatoolbox vibecheck datatops --quiet\n",
+    "!pip install numpy xarray scipy scikit-learn matplotlib seaborn tqdm vibecheck datatops --quiet\n",
+    "!pip install rsatoolbox==0.1.5 --quiet\n",
     "\n",
     "from vibecheck import DatatopsContentReviewContainer\n",
     "def content_review(notebook_section: str):\n",
@@ -1003,7 +1004,7 @@
     "n_neurons = 2\n",
     "stimulus_idx = 0,1 # choose two stimuli\n",
     "#################################################\n",
-    "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
+    "raise NotImplementedError(\"Student exercise: complete Euclidean distance calculation by the formula provided above\")\n",
     "#################################################\n",
     "b_j = clean_dataset[n_neurons].loc[stimulus_idx[0]].values # select the stimulus response\n",
     "b_k = clean_dataset[n_neurons].loc[stimulus_idx[1]].values\n",
@@ -1368,9 +1369,9 @@
     "    projected_dist[n_neurons]=[]\n",
     "    for m_dims in m_dims_list:\n",
     "        #################################################\n",
-    "        raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
+    "        raise NotImplementedError(\"Student exercise: generate matrix A which projects from dimensionality neurons's amount to d-dimensional space\")\n",
     "        #################################################\n",
-    "        A = ...\n",
+    "        A = np.random.normal(loc=..., scale=..., size=(..., ...))\n",
     "        A *= np.sqrt(1/m_dims)\n",
     "        transformed_data = (data.values @ A)\n",
     "        transformed_data = np2xr(transformed_data, coords={'stim': data.stim.values, 'neuron': np.arange(m_dims)})\n",

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/instructor/W1D3_Tutorial1.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/instructor/W1D3_Tutorial1.ipynb
@@ -113,8 +113,9 @@
    "source": [
     "# @title Install dependencies\n",
     "\n",
-    "!pip install -q ipympl ipywidgets mpl_interactions[\"jupyter\"] rsatoolbox torchlens\n",
+    "!pip install -q ipympl ipywidgets mpl_interactions[\"jupyter\"] torchlens\n",
     "!pip install -q graphviz\n",
+    "!pip install rsatoolbox==0.1.5\n",
     "\n",
     "# To install jupyter-matplotlib (ipympl) via pip\n",
     "!pip install -q torchlens\n",

diff --git a/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/instructor/W1D3_Tutorial2.ipynb b/tutorials/W1D3_ComparingArtificialAndBiologicalNetworks/instructor/W1D3_Tutorial2.ipynb
@@ -90,7 +90,8 @@
    "source": [
     "# @title Install and import feedback gadget\n",
     "\n",
-    "!pip install torch torchvision matplotlib numpy scikit-learn rsatoolbox scipy vibecheck --quiet\n",
+    "!pip install torch torchvision matplotlib numpy scikit-learn scipy vibecheck --quiet\n",
+    "!pip install rsatoolbox==0.1.5 --quiet\n",
     "\n",
     "from vibecheck import DatatopsContentReviewContainer\n",
     "def content_review(notebook_section: str):\n",
@@ -1269,7 +1270,7 @@
    "source": [
     "# @title Representational Path\n",
     "\n",
-    "imgs, labels = next(iter(test_loader)) #grab 500 samples from the test set\n",
+    "imgs, labels = next(iter(test_loader))\n",
     "\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "\n",
@@ -1330,7 +1331,6 @@
    "source": [
     "# @title Test your idea by executing this cell!\n",
     "\n",
-    "#imgs, labels = next(iter(test_loader)) #grab 500 samples from the test set\n",
     "imgs, labels = sample_images(test_loader, n=5)\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "\n",
@@ -1476,45 +1476,24 @@
     "execution": {}
    },
    "source": [
-    "## Coding Exercise 1: Dimensionality reduction visualization\n",
+    "## Dimensionality reduction visualization\n",
     "\n",
-    "Use MDS to visualize the changes in the representational geometry across the layers of this network in response to original MNIST test images."
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {
-    "colab_type": "text",
-    "execution": {}
-   },
-   "source": [
-    "```python\n",
-    "#################################################\n",
-    "# Fill out and remove\n",
-    "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
-    "#################################################\n",
-    "\n",
-    "return_layers = ['input', 'conv1', 'conv2', 'fc1', 'fc2']\n",
-    "imgs, labels = sample_images(test_loader, n=50) # grab 500 samples from the test set\n",
-    "model_features = ...\n",
-    "\n",
-    "plot_dim_reduction(model_features, labels, transformer_funcs =['MDS'])\n",
-    "\n",
-    "```"
+    "Now, we are going to use MDS to visualize the changes in the representational geometry across the layers of this network in response to the original MNIST test images."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
+    "cellView": "form",
     "execution": {}
    },
    "outputs": [],
    "source": [
-    "# to_remove solution\n",
+    "# @title Plot the representational geometry changes\n",
     "\n",
     "return_layers = ['input', 'conv1', 'conv2', 'fc1', 'fc2']\n",
-    "imgs, labels = sample_images(test_loader, n=50) # grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) # grab 50 samples from the test set\n",
     "model_features = extract_features(model_robust, imgs.to(device), return_layers)\n",
     "\n",
     "plot_dim_reduction(model_features, labels, transformer_funcs =['MDS'])"
@@ -1558,7 +1537,7 @@
    },
    "outputs": [],
    "source": [
-    "imgs, labels = sample_images(test_loader, n=50) #grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) #grab 50 samples from the test set\n",
     "\n",
     "model_features = extract_features(model, imgs.to(device), return_layers='all')\n",
     "model_features_robust = extract_features(model_robust, imgs.to(device), return_layers='all')\n",
@@ -1585,7 +1564,7 @@
    },
    "outputs": [],
    "source": [
-    "imgs, labels = sample_images(test_loader, n=50) #grab 500 samples from the test set\n",
+    "imgs, labels = sample_images(test_loader, n=50) #grab 50 samples from the test set\n",
     "\n",
     "adv_imgs = generate_adversarial(model, imgs.to(device), labels.to(device), eps)\n",
     "\n",
@@ -1625,35 +1604,9 @@
     "execution": {}
    },
    "source": [
-    "## Coding Exercise 2: Compare representational paths\n",
-    "\n",
-    "Train another instance of the model on the original MNIST data for only one epoch with a different seed. Then compare the representational paths of the two instances that are both trained on the original MNIST with one another. Display the paths in blue and cyan colors."
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {
-    "colab_type": "text",
-    "execution": {}
-   },
-   "source": [
-    "```python\n",
-    "#################################################\n",
-    "## TODO for students: fill in the missing variables ##\n",
-    "# Fill out function and remove\n",
-    "raise NotImplementedError(\"Student exercise: fill in the missing variables\")\n",
-    "#################################################\n",
+    "## Compare representational paths\n",
     "\n",
-    "args = build_args()\n",
-    "torch.manual_seed(args.seed+1)\n",
-    "args.epochs = 1\n",
-    "#build_model\n",
-    "model_new_seed = Net().to(device)\n",
-    "\n",
-    "optimizer = optim.Adadelta(...)\n",
-    "train_model(args, model_new_seed, optimizer)\n",
-    "\n",
-    "```"
+    "Train another instance of the model on the original MNIST data for only one epoch with a different seed (the basic model is trained for one epoch too). Then compare the representational paths of the two instances that are both trained on the original MNIST with one another. Display the paths in blue and cyan colors."
    ]
   },
   {
@@ -1664,8 +1617,6 @@
    },
    "outputs": [],
    "source": [
-    "# to_remove solution\n",
-    "\n",
     "args = build_args()\n",
     "torch.manual_seed(args.seed+1)\n",
     "args.epochs = 1\n",