diff --git a/biography.html b/biography.html index 134a519..ecc75a6 100644 --- a/biography.html +++ b/biography.html @@ -62,4 +62,11 @@ }); }); + +
+

Biography

+

This is my biography

+
+ + \ No newline at end of file diff --git a/images/EC_white.png b/images/EC_white.png new file mode 100644 index 0000000..f2fa038 Binary files /dev/null and b/images/EC_white.png differ diff --git a/images/PF6_white.png b/images/PF6_white.png new file mode 100644 index 0000000..a3b3198 Binary files /dev/null and b/images/PF6_white.png differ diff --git a/images/bayesopt.png b/images/bayesopt.png new file mode 100644 index 0000000..6e21ef8 Binary files /dev/null and b/images/bayesopt.png differ diff --git a/images/bayesoptsmalltrans.png b/images/bayesoptsmalltrans.png new file mode 100644 index 0000000..e3c4d29 Binary files /dev/null and b/images/bayesoptsmalltrans.png differ diff --git a/images/mean-acqf-EC-PF6.png b/images/mean-acqf-EC-PF6.png new file mode 100644 index 0000000..91c087e Binary files /dev/null and b/images/mean-acqf-EC-PF6.png differ diff --git a/images/molsets_diagram.png b/images/molsets_diagram.png new file mode 100644 index 0000000..03f9975 Binary files /dev/null and b/images/molsets_diagram.png differ diff --git a/index.html b/index.html index 2479d77..f3af91d 100644 --- a/index.html +++ b/index.html @@ -48,7 +48,7 @@ -
+
Main Image @@ -56,7 +56,7 @@
-
+
Main Image @@ -64,39 +64,35 @@
-
+
-
+

Cooper Tezak

-
-
+
-
-

PhD Student, CU Boulder

+

PhD Candidate, CU Boulder

Atomic Broker

-
-
-
+
+ -
+ -
+ -
@@ -129,7 +125,7 @@
-
+
+
Battery Electrolyte Composition Optimization
I have recently begun working with Chunmei Ban - on optimizing experimental electrolyte composition for Na-ion batteries. I lead the effort to develop a Bayesian optimization framework - that uses a gaussian process to estimate the charge capacity at a given cycle based on the mass weights of electrolyte components. + on optimizing experimental electrolyte composition for Na-ion batteries. I have succesfully implemented + information theoretic and simple expected improvement acquisiton functions on top of a gaussian process + surrogate model to efficiently explore compositions experimentally. However, the design space for electrolyte compositions + is huge and experiments are expensive. I am currently working on incorporating molecular features into the model using + graph neural networks to featurize molecular structures and attention mechanisms to create mixture level representations + from the constituent molecular features. + [7] + The architecture of the model is summarized as

+ Molsets Architecture

@@ -139,6 +146,13 @@ to have less occupied d-orbitals that were capable of strongly reducing the \( N_2 \) triple bond. However, all materials were inevitably limited by the desorption of \( NH_3* \) from the surface.

+ Bayesian Optimization +

+ Bayesian Optimization model for battery electrolyte compositions. X and Y axes are fractions of chemicals + EC and NaPF6. The left panel shows the mean of the model and the right shows the + acquisiton function value, which is the + log expected improvement + .