Create environment using environment.yml.

Details to run are provided with each experiment. All experiments are logged using logged using wandb. If you want to use some other logger or save results locally modify the logging part of the code accordingly.

FFHQ: Results

Old → Young

(a) Static subset	(b) Dynamic subset (Euler 100 steps)

_{Figure 1. FFHQ old → young translation using (a) static and (b) dynamic subset selection. The dynamic version is evaluated using Euler integration (100 steps).}

---

Young → Old

(a) Static subset	(b) Dynamic subset (Euler 100 steps)

_{Figure 2. FFHQ young → old translation using (a) static and (b) dynamic subset selection. The dynamic version is evaluated using Euler integration (100 steps).}

---

Woman → Man

(a) Static subset	(b) Dynamic subset (Euler 100 steps)

_{Figure 3. FFHQ woman → man translation using (a) static and (b) dynamic subset selection. The dynamic version is evaluated using Euler integration (100 steps).}

---

Man → Woman

(a) Static subset	(b) Dynamic subset (Euler 100 steps)

_{Figure 4. FFHQ man → woman translation using (a) static and (b) dynamic subset selection. The dynamic version is evaluated using Euler integration (100 steps).}

MNIST → EMNIST: Results

Static	Dynamic

_{Figure 5. MNIST → EMNIST. The dynamic version is evaluated using Euler integration (100 steps).}

In order to understand the principles behind paper, please start with static subset alignment between toy point distributions start with this notebook and dynamic with this notebook.

Static example c=2	Dynamic example c=2
Evolution of dynamic potential with time between two uniform distributions

Evolution of dynamic potential with time between two bimodal uniform distributions