change recommendation from scipy.optimize.minimize to skopt.optimizer.gp_minimize -- works much better (and faster)!

Author: jeremymanning

content/assignments/Assignment_2:Search_of_Associative_Memory_Model/README.md

@@ -110,7 +110,7 @@ You will fit **eight parameters** to optimize the match to human recall data:
 7. $m_{1_{\text{max}}}$: maximum number of *contextual* association cueing failures
 8. $m_{2_{\text{max}}}$: maximum number of *episodic* association cueing failures
 
-You can choose any approach you wish to fit these parameters.  My "recommended" approach is to use [scipy.optimize.minimize](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html) to minimize the mean squared error between the point-by-point observed vs. model-predicted values for the following behavioral curves:
+You can choose any approach you wish to fit these parameters.  My "recommended" approach is to use [skopt.optimizer.gp_minimize](https://scikit-optimize.github.io/stable/modules/generated/skopt.optimizer.gp_minimize.html#skopt.optimizer.gp_minimize) to minimize the mean squared error between the point-by-point observed vs. model-predicted values for the following behavioral curves:
   - $p(\text{first recall})$: probability of recalling each item **first** as a function of its *presentation position*
   - $p(\textit{recall})$: probability of recalling each item at *any* output position as a function of its presentation position
   - lag-CRP: probability of recalling item $i$ given that item $j$ was the previous recall, as a function of $lag = i - j$.
@@ -139,7 +139,7 @@ You can use the [example notebook](https://contextlab.github.io/memory-models-co
 - To help with computing mean squared error, it will be useful to have a function that takes in a dataset as input and returns a vector comprising each of these curves, for each list length and presentation rate, concatenated together into a single vector.
 
 ### **Step 4: Fit Model Parameters**
-- To compute mean squared error for a given set of model parameters, use the function you wrote above to compute the concatenated behavioral curves for the *observed recalls* and the *model-predicted recalls*.  The average squared point-by-point difference between the vectors is the mean squared error.  You'll want to set up [scipy.optimize.minimize](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html) to find the set of model parameters that minimizes the mean squared error between the observed and predicted curves, using only the training dataset.
+- To compute mean squared error for a given set of model parameters, use the function you wrote above to compute the concatenated behavioral curves for the *observed recalls* and the *model-predicted recalls*.  The average squared point-by-point difference between the vectors is the mean squared error.  You'll want to set up [skopt.optimizer.gp_minimize](https://scikit-optimize.github.io/stable/modules/generated/skopt.optimizer.gp_minimize.html#skopt.optimizer.gp_minimize) to find the set of model parameters that minimizes the mean squared error between the observed and predicted curves, using only the training dataset.
 - Importantly, you should use the same parameters across all trials and experimental conditions.  You're fitting the *average* performance, not data from individual trials or participants.
 
 ### **Step 5: Generate Key Plots**

content/assignments/Assignment_2:Search_of_Associative_Memory_Model/sam_assignment_template.ipynb

@@ -203,7 +203,7 @@
         "Other tasks:\n",
         "  - Fit params to [Murdock (1962) dataset](https://github.com/ContextLab/memory-models-course/tree/main/content/assignments/Assignment_2%3ASearch_of_Associative_Memory_Model/Murd62%20data) that you downloaded with the `load_data` function.\n",
         "     - You'll need to define a \"loss\" function.  I suggest computing MSE for one or more behavioral curves, computed for a subset of the Murdock (1962) participants/lists\n",
-        "     - I suggest using [scipy.optimize.minimize](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html) to estimate the model parameters.\n",
+        "     - I suggest using [skopt.optimizer.gp_minimize](https://scikit-optimize.github.io/stable/modules/generated/skopt.optimizer.gp_minimize.html#skopt.optimizer.gp_minimize) to estimate the model parameters.\n",
         "  - Create observed/predicted plots for held-out data:\n",
         "     - p(first recall)\n",
         "     - p(recall)\n",

content/assignments/Assignment_3:Context_Maintenance_and_Retrieval_Model/README.md

@@ -120,7 +120,7 @@ Fit the model to the following curves and measures from the Polyn et al. (2009)
 There are several possible ways to accomplish this.  My recommended approach is:
 1. Split the dataset into a training set and a test set
 2. Compute the above curves/measures for the training set and concatenate them into a single vector
-3. Use [scipy.optimize.minimize](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html#scipy.optimize.minimize) to find the set of model parameters that minimizes the mean squared error between the observed curves and the CMR-estimated curves (using the given parameters).
+3. Use [skopt.optimizer.gp_minimize](https://scikit-optimize.github.io/stable/modules/generated/skopt.optimizer.gp_minimize.html#skopt.optimizer.gp_minimize) to find the set of model parameters that minimizes the mean squared error between the observed curves and the CMR-estimated curves (using the given parameters).
 4. Compare the observed performance vs. CMR-estimated performance (using the best-fitting parameters) for the test data
 
 

content/assignments/Assignment_3:Context_Maintenance_and_Retrieval_Model/cmr_assignment_template.ipynb

@@ -3,8 +3,8 @@
     {
       "cell_type": "markdown",
       "metadata": {
-        "id": "view-in-github",
-        "colab_type": "text"
+        "colab_type": "text",
+        "id": "view-in-github"
       },
       "source": [
         "<a href=\"https://colab.research.google.com/github/ContextLab/memory-models-course/blob/main/content/assignments/Assignment_3%3AContext_Maintenance_and_Retrieval_Model/cmr_assignment_template.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
@@ -113,19 +113,13 @@
         "\n",
         "data = load_data()"
       ]
-    },
-    {
-      "cell_type": "code",
-      "source": [],
-      "metadata": {
-        "id": "IjVqOOsZEM4q"
-      },
-      "id": "IjVqOOsZEM4q",
-      "execution_count": null,
-      "outputs": []
     }
   ],
   "metadata": {
+    "colab": {
+      "include_colab_link": true,
+      "provenance": []
+    },
     "kernelspec": {
       "display_name": "memory-course",
       "language": "python",
@@ -142,12 +136,8 @@
       "nbconvert_exporter": "python",
       "pygments_lexer": "ipython3",
       "version": "3.11.0"
-    },
-    "colab": {
-      "provenance": [],
-      "include_colab_link": true
     }
   },
   "nbformat": 4,
   "nbformat_minor": 5
-}
+}