# Computation Validation

To guarantee the integrity of geo-distributed GPU contributions, we introduce a computation validation mechanism. This mechanism uses two techniques.

* **Adaptive verification based on computation redundancy**. We replicate computations on multiple geo-distributed GPUs and compare their computation results. Such computation redundancy can consume extra AI compute resources. To minimize redundancy, we use random auditing: the redundancy is randomly created on random GPUs. Such a method creates an unpredictable verification system that intimidates dishonesty.
* **Prediction-guided verification**. For the use cases of AI fine-tuning or training, we employ a continuous validation method. In particular, we analyze data patterns and the trend of model convergence during training. Then, we predict the model parameter evolution using empirical models \[1], and compare the prediction against the computation results. This dynamic validation mechanism can identify inconsistencies in parameter updates

Reference:

1. L. Liang, R. Chen, H. Chen, Y. Xia, K. Park, B. Zang, and H. Guan, “SLAQ: Quality-Driven Scheduling for Distributed Machine Learning,” in *ACM Symposium on Cloud Computing (SoCC)*, 2017


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.yottalabs.ai/technology/computation-validation.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.