Home Knowledge Base Perceiver

Perceiver is a general-purpose transformer architecture that uses cross-attention to project arbitrary-size inputs into a fixed-size latent array — decoupling the computational cost from input size so that a 100K-pixel image, a 50K-token audio clip, and a 10K-point cloud all get processed through the same small latent bottleneck (e.g., 512 latent vectors), enabling a single architecture to handle any modality without modality-specific design choices.

What Is Perceiver?

Architecture

StepOperationInputOutputComplexity
1. Cross-AttentionLatent queries attend to inputLatent: M × d, Input: N × d_inM × d (latent updated)O(M × N)
2. Self-AttentionLatent self-attention (multiple blocks)M × dM × d (refined)O(M²) per block
3. Repeat (optional)Additional cross-attention + self-attentionUpdated latent + original inputM × d (further refined)O(M × N + M²)
4. DecodeTask-specific output (class token, etc.)M × dTask outputO(M)

Key Insight: The Latent Bottleneck

PropertyStandard TransformerPerceiver
Self-attention costO(N²) — depends on input sizeO(M²) — depends on latent size (fixed)
Input flexibilityFixed tokenization per modalityAny byte array, any modality
ScalabilityCost grows quadratically with inputCost fixed regardless of input size
Architecture per modalityDifferent: ViT for images, BERT for textSame architecture for everything

Example: M=512 latents, N=50,000 input elements:

Modality Flexibility

ModalityInput RepresentationSame Perceiver Architecture
ImagesPixel array (H×W×C) with positional encoding
AudioRaw waveform or spectrogram
Point Clouds3D coordinates (N×3)
VideoPixel frames (T×H×W×C)
TextToken embeddings
MultimodalConcatenate all modalities as one input array

Perceiver is the universal perception architecture — using cross-attention to a fixed-size latent array to decouple computational cost from input size and modality, enabling a single unchanged architecture to process images, audio, video, point clouds, and multimodal inputs with O(M²) self-attention cost regardless of whether the input has 1,000 or 1,000,000 elements, pioneering the movement toward truly modality-agnostic deep learning.

perceiverfoundation model

Explore 500+ Semiconductor & AI Topics

From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.