published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the "Taverne" license above, please follow below link for the End User

The number of systems that are investigated for computation in the physical domain has increased substantially in the recent past. Optical and photonic systems have drawn high interest due to their potential for carrying out energy-efficient linear operations and perceived advantages in latency and general computation speed. One of the main challenges remains to scale up integrated photonic designs to integration densities required for meaningful computation, in particular for matrix-vector multiplications. To address upscaling for photonic computing, here we propose an on-chip scheme for dimension reduction of the input data using random scattering. Exploiting tailored disorder allows us to reduce the incoming dimensionality by more than an order of magnitude, which a shallow subsequent network can use to perform image recognition tasks with high accuracy.

Integrated optical processing networks enable high computation speeds combined with low energy consumption. We present here a scheme for dimension reduction for optical neural networks, by orders of magnitudes, while still reaching high classification accuracies.