Convolutional Autoencoder for Mask Shape Data Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The integrated circuit manufacturing industry faces significant challenges in efficiently processing and storing large volumes of data for mask patterns due to the limitations of conventional compression techniques, which result in excessive computation time and significant data loss.
Innovation Solution
The use of a convolutional autoencoder trained with machine learning algorithms to compress shape data for electronic designs, specifically tuning parameters to retain important information based on design rules, allowing for efficient encoding and decoding of mask designs while minimizing data loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional compression techniques are used to compress shape data for mask patterns, then data volume is reduced, but computation time increases excessively and data loss becomes significant
Solution Approach 1:
The patent transforms the compression approach by changing from conventional compression algorithms to a neural network-based autoencoder system. The key parameter change is using learned representations (embeddings) that capture essential shape characteristics in a compressed form, achieving both data reduction and fast decompression. The neural network is trained to preserve critical geometric information while discarding redundant details, resolving the trade-off between compression ratio and computation time.
Solution Approach 2:
The patent replaces conventional mechanical/computational compression algorithms with a neural network-based system. The autoencoder learns to compress shape data through training on example mask patterns, substituting traditional algorithmic approaches with learned transformations. This substitution enables the system to achieve compression without the excessive computation time associated with conventional techniques, as the neural network performs compression through learned feature extraction rather than iterative algorithmic processing.
2Quantity of substance
If conventional compression techniques are used to compress shape data for mask patterns, then data volume is reduced, but data loss becomes significant
Solution Approach 1:
The patent changes the compression paradigm by using neural network embeddings that learn to represent shapes in a compressed latent space. The autoencoder is trained to preserve critical geometric features (such as corner positions, line segments, and spatial relationships) while compressing the data. This learned representation approach maintains data fidelity better than conventional compression, as it focuses on preserving semantically important features rather than uniformly compressing all data points.
Solution Approach 2:
The patent extracts and preserves only the essential geometric features of mask patterns through the neural network's latent representation. The autoencoder learns to identify and retain critical shape characteristics (corners, edges, spatial relationships) while discarding redundant information. This selective extraction of important features enables significant data compression without significant data loss, as the compressed representation maintains the essential geometric information needed for accurate pattern reproduction.
3Productivity
If neural network autoencoder is used to compress shape data, then processing efficiency is improved, but system complexity increases
Solution Approach 1:
The patent applies preliminary action by training the neural network autoencoder offline on a dataset of example mask patterns before actual compression is needed. During this training phase, the system learns optimal compression representations and geometric feature preservation strategies. Once trained, the model can rapidly compress new shape data without requiring complex real-time processing. This preliminary training step separates the complex learning process from the actual compression operation, improving processing efficiency while managing system complexity through offline preparation.
Data Source
AI summary
Methods and systems for compressing shape data for a set of electronic designs include inputting a set of shape data, where the shape data comprises mask designs. A convolutional autoencoder encodes the set of shape data, where the encoding compresses the set of shape data to produce a set of encoded shape data. The convolutional autoencoder is tuned for increased accuracy of the set of encoded shape data based on design rules for the set of electronic designs. The convolutional autoencoder comprises a set of parameters comprising weights, and the convolutional autoencoder has been trained to retain important information needed, based on the design rules for the set of electronic designs.