Constructive criticism from a SR&ED consultancy firm of our past claims
Source: Notion | Last edited: 2025-03-20 | ID: 1bc2d2dc-3ef...
Although we certainly hope to have the opportunity to work with Eon Labs, the following feedback is complementary and can hopefully serve to improve future claims wherever you ultimately end up having your SR&ED prepared.
Feedback from our technical writer:
The claim has several issues, particularly regarding statements about the non-existence of similar technologies that need revision.
https://www.sciencedirect.com/science/article/abs/pii/S0957417423020225
This article’s bibliography references earlier versions of this approach, so you should avoid claiming generic originality in SR&ED narratives. Instead, focus on your specific implemented divergences.
When discussing technological uncertainties, you must clearly explain why they qualify as such. This explanation was needed in the last two claimed years. While the claims were accepted as filed, it’s better to be more thorough going forward. Remember that using TensorFlow APIs alone doesn’t qualify as developing new models—unless you’re extending their functionality. Similarly, while using Kalman filters is promising, they’ve already been used with CNNs. My background is in Computer Science, not specifically neural networks, though I have implemented a multiple GAN application for image processing and written several SR&ED claims for neural network applications. A common weakness in these claims is assuming that merely using TensorFlow APIs qualifies as SR&ED. The real SR&ED value comes from experimental datasets and optimizing their ingestion into the neural network model. While your narratives cover this, they contain too much repetition that should be reduced in future submissions.
I hope that this feedback will prove valuable for your team. And thank you for considering our firm.
The reviewing consultant’s background:
My background is in Computer Science per se. It is not in neural network systems. However, I have recently implemented a multiple GAN application for image processing and I have written several sred claims for neural network applications. I am aware of what is problematic about some neural network claims and the conceptual requirements of description in sred narratives. I have written 100s of sred claims over the last 20 years on a part time freelance basis. During that period, I have also been involved in my own software companies & have patents.