🎯 Explore Trade-offs

Focus question: How do my objectives balance against each other, and where are the best trade-offs?

Best Trade-off Points

(Pareto solutions)

Pareto solutions

What it is: A tabular summary of experiments that lie on the Pareto front of multiple objectives.

How to use it:

• Scan rows to compare performance trade-offs between objectives.
• Identify solutions that balance priorities across outputs rather than optimizing just one.
• Use as a shortlist of candidates for practical decision-making.

Why it matters:

• Focuses attention on the most efficient trade-off options.
• Supports informed choices when objectives conflict.

Limitations:

• Only highlights trade-offs that are captured by the current model/data.
• Does not express preferences—requires external weighting or judgment.
• Does not consider uncertainty and may contain experiments already tested.

	input_1	input_2	input_3	input_4	input_5	input_6	input_7	input_8	output_1	output_2
0	99.0	NMP	9.62	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.59	0.0 ± 0.0	99.982 ± 0.27
1	102.0	DMAc	9.70	1.0	DPPP	0.016	(TMS)3SiH	0.11	87.518 ± 18.56	0.001 ± 0.22

Table 1: Table summarizing Pareto-optimal solutions, where no objective can be improved without degrading another. This view highlights the set of non-dominated experiments for decision-making in multi-objective optimization.

Pareto Front

(Multi-objective Scatter)

Multi-objective Scatter

What it is: A graphical view of the Pareto front: the curve (or surface) of best trade-offs between competing objectives.

How to use it:

• Look for the frontier points—those lying on the outer edge of the scatter.
• Evaluate where along the frontier your preferred balance lies.
• Use to understand the cost of gaining one objective in terms of another.

Why it matters:

• Makes trade-offs visually clear and intuitive.
• Supports negotiation between competing design goals.
• Highlights diminishing returns (when pushing one metric hurts another sharply).

Limitations:

• Hard to visualize beyond two or three objectives.
• Does not encode uncertainty—frontier may shift as data/model updates.

(a) output_1-vs-output_2

Figure 1: 2D scatter plots visualizing the Pareto front, showing the best trade-offs between objectives. These plots reveal how improving one metric impacts another and help select balanced solutions. Note: difficult to interpret with more than 2–3 objectives.

Independently Optimized Outputs

(Single-objective Optima)

Single-objective Optima

What it is: The best result achievable for each objective if optimized independently.

How to use it:

• Use to benchmark the maximum possible value for each metric.
• Compare to Pareto solutions to see how much is sacrificed for balance.
• Helpful for stakeholders focused on one objective above all.

Why it matters:

• Clarifies the theoretical upper or lower bounds for each output.
• Helps quantify the trade-off gap when moving from single-objective to multi-objective optimization.

Limitations:

• Not feasible to achieve all independent optima simultaneously.
• Can mislead if taken out of context without trade-off analysis.

	Output Optimized	input_1	input_2	input_3	input_4	input_5	input_6	input_7	input_8	output_1	output_2
0	output_1	102.0	DMAc	9.70	1.0	DPPP	0.016	(TMS)3SiH	0.11	87.518 ± 18.56	0.001 ± 0.22
1	output_2	103.0	NMP	9.95	1.0	L59 | N-XantPhos	0.015	(TMS)3SiH	0.60	0.0 ± 0.0	99.982 ± 0.49

Table 2: Table of best solutions found when optimizing each output individually. This isolates the maximum potential of each objective but ignores trade-offs with others.

Compare Output Trade-offs

(Parallel coordinates plot)

Parallel coordinates plot

What it is: A multi-dimensional plot where each axis represents an output, and each line is a candidate solution spanning all objectives.

How to use it:

• Trace lines across axes to see how each solution performs across objectives.
• Look for lines staying consistently high (balanced solutions).
• Spot crossing lines where improving one objective worsens another.

Why it matters:

• Allows simultaneous comparison of many objectives—beyond 2D Pareto plots.
• Reveals global trade-off structure across outputs.
• Supports decision-making by highlighting balanced vs extreme solutions.

Limitations:

• Can become cluttered with many points or objectives.
• Scaling of axes can strongly affect visual interpretation.
• Harder for non-experts to read compared to Pareto front plots.