Familiarity with Flux.jlUnderstanding of convolutional layersBasic knowledge of recurrent networks
新手友善度1-100 的估計分數,表示該議題對首次貢獻者的友善程度。
35
研究方向
Refer to the linked MXNet implementation for reference. Start by examining existing convolutional layers in Flux (e.g., Conv, ConvTranspose) and recurrent cells (e.g., LSTM, GRU). Implement a ConvRNN cell similar to the MXNet version, then combine multiple cells to form ConvLSTM. Ensure compatibility with Flux's automatic differentiation. Open a draft PR early for feedback.