The paper introduces wavelet scattering networks, which compute translation-invariant image representations that are stable to deformations and preserve high-frequency information for classification. The network cascades wavelet transform convolutions with non-linear modulus and averaging operators. The first layer outputs SIFT-type descriptors, while subsequent layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. The scattering representation of stationary processes incorporates higher-order moments, allowing for discrimination of textures with the same Fourier power spectrum. State-of-the-art classification results are achieved for handwritten digit recognition and texture discrimination using a Gaussian kernel SVM and a generative PCA classifier. The paper also discusses the deformation stability of the scattering transform, the properties of scattering coefficients, and the optimization of the network architecture. Additionally, it explores the use of a cosine transform to decorrelate scattering coefficients and a fast computational algorithm for scattering computations. Finally, the paper presents classification results using scattering vectors with PCA and SVM classifiers.The paper introduces wavelet scattering networks, which compute translation-invariant image representations that are stable to deformations and preserve high-frequency information for classification. The network cascades wavelet transform convolutions with non-linear modulus and averaging operators. The first layer outputs SIFT-type descriptors, while subsequent layers provide complementary invariant information that improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification. The scattering representation of stationary processes incorporates higher-order moments, allowing for discrimination of textures with the same Fourier power spectrum. State-of-the-art classification results are achieved for handwritten digit recognition and texture discrimination using a Gaussian kernel SVM and a generative PCA classifier. The paper also discusses the deformation stability of the scattering transform, the properties of scattering coefficients, and the optimization of the network architecture. Additionally, it explores the use of a cosine transform to decorrelate scattering coefficients and a fast computational algorithm for scattering computations. Finally, the paper presents classification results using scattering vectors with PCA and SVM classifiers.