diff --git a/examples/15_pysindy_lectures.ipynb b/examples/15_pysindy_lectures.ipynb
index a74178baa..c19b1be28 100644
--- a/examples/15_pysindy_lectures.ipynb
+++ b/examples/15_pysindy_lectures.ipynb
@@ -1484,7 +1484,7 @@
    "source": [
     "# Part 4: How to choose a regularizer and a sparse regression algorithm?\n",
     "This table summarizes the optimizers available in PySINDy. Note that TrappingSR3 and SINDyPI are both geared for very specific dynamical systems, so we will only investigate the remaining optimizers.\n",
-    "![title](optimizer_summary.jpg)\n",
+    "![title](data/optimizer_summary.jpg)\n",
     "### Okay so how do you choose between, for instance, using the $l_0$ and $l_1$ regularizers? And once that choice is made, which algorithm should you use to solve the problem?\n",
     "#### Advantages and disadvantages of $l_0$: \n",
     "Using the $l_0$ norm typically produces sparser solutions than using the $l_1$ norm. This tends to further lead to higher performance and more stable models, since there are no small-coefficient terms that can become active with new initial conditions or parameter regimes. The downside is that the $l_0$ norm transforms the SINDy regression into a nonconvex problem, for which only local convergence guarantees can be provided. <br>\n",
@@ -2204,7 +2204,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.9"
+   "version": "3.7.4"
   },
   "toc": {
    "base_numbering": 1,
diff --git a/docs/JOSS2/optimizer_summary.jpg b/examples/data/optimizer_summary.jpg
similarity index 100%
rename from docs/JOSS2/optimizer_summary.jpg
rename to examples/data/optimizer_summary.jpg