René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Cannot apply element-wise operator Add","articleBody":"I examined DeepNet sample code referenced by (http://www.deepml.net/model.html)\r\n\r\nHowever, I executed my code and got System.Exception. Additional information was as follows:\r\n\r\n```\r\ncannot apply element-wise operation Add to unequal shapes [\"nHidden\"; \"nBatch\"] and [\"nHidden\"; \"nHidden\"]\r\n```\r\nThe exception occurred at `let hiddenAct = hiddenWeights .* input.T + hiddenBias`\r\nI expect `hiddenBias` will be `[nHidden; nBatch]` shapes, but `[nHidden; nHidden]`.\r\n\r\nMy complete code is as follows:\r\n\r\n```fsharp\r\nopen Tensor\r\nopen Datasets\r\nopen SymTensor\r\n\r\n[\u003cEntryPoint\u003e]\r\nlet main argv = \r\n    printfn \"%A\" argv\r\n\r\n    let a = HostTensor.init [7L; 5L] (fun [|i; j|] -\u003e 5.0 * float i + float j) \r\n\r\n    /// MINST Dataset\r\n    let mnist = Mnist.load(__SOURCE_DIRECTORY__ + \"../../MNIST\") 0.0 |\u003e TrnValTst.toHost\r\n    \r\n    printfn \"MNIST training set: images have shape %A and labels have shape %A\" mnist.Trn.All.Input.Shape mnist.Trn.All.Target.Shape   \r\n    printfn \"MNIST test set    : images have shape %A and labels have shape %A\" mnist.Tst.All.Input.Shape mnist.Tst.All.Target.Shape\r\n\r\n    /// Definition NeuralNetModel\r\n    let mb = ModelBuilder\u003csingle\u003e \"NeuralNetModel\"\r\n\r\n    // Definition symbol\r\n    let nBatch  = mb.Size \"nBatch\"\r\n    let nInput  = mb.Size \"nInput\"\r\n    let nClass  = mb.Size \"nClass\"\r\n    let nHidden = mb.Size \"nHidden\"\r\n\r\n    // Model paramaters\r\n    let hiddenWeights = mb.Param (\"hiddenWeights\", [nHidden; nInput])\r\n    let hiddenBias    = mb.Param (\"hiddenBias\"   , [nHidden])\r\n    let outputWeights = mb.Param (\"outputWeights\", [nClass; nHidden])\r\n\r\n    // Model variables\r\n    let input  = mb.Var\u003csingle\u003e \"Input\"  [nBatch; nInput]\r\n    let target = mb.Var\u003csingle\u003e \"Target\" [nBatch; nClass]\r\n\r\n    // Generating model\r\n    mb.SetSize nInput mnist.Trn.All.Input.Shape.[1]\r\n    mb.SetSize nClass mnist.Trn.All.Target.Shape.[1]\r\n    mb.SetSize nHidden 100L\r\n\r\n    let mi = mb.Instantiate DevHost\r\n\r\n    // Definition model action in input -\u003e hidden\r\n    let hiddenAct = hiddenWeights .* input.T + hiddenBias // \u003c--------- Exception occurrs!!!\r\n    let hiddenVal = tanh hiddenAct\r\n\r\n    // Definition model action in hidden -\u003e output\r\n    let outputAct = outputWeights .* hiddenVal\r\n    let classProb = exp outputAct / Expr.sumKeepingAxis 0 (exp outputAct)\r\n\r\n    // Loss function\r\n    let smplLoss = - Expr.sumAxis 0 (target.T * log classProb)\r\n    let loss     = Expr.mean smplLoss\r\n\r\n    // Compile\r\n    let lossFn   = mi.Func loss |\u003e arg2 input target\r\n\r\n    // Initialization with seed\r\n    mi.InitPars 123\r\n\r\n    // test\r\n    let tstLossUntrained = lossFn mnist.Tst.All.Input mnist.Tst.All.Target |\u003e Tensor.value\r\n\r\n    printfn \"Test loss (untrained): %.4f\" tstLossUntrained\r\n\r\n    System.Console.ReadKey() |\u003e ignore\r\n    0 // exit code\r\n```\r\n\r\nMy environment are as follows:\r\n\r\n* Windows 7 (64bit)\r\n* Visual Studio 2015\r\n* Installed DeepNet via Nuget\r\n* Installed FSharp.Core(F# 4.1) via Nuget\r\n\r\nI'm sorry if I'm misunderstanding about your sophisticated library.\r\nCould you please let me know how to fix this problem?\r\n \r\n","author":{"url":"https://github.com/kuroyakov","@type":"Person","name":"kuroyakov"},"datePublished":"2017-05-22T03:18:11.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":4},"url":"https://github.com/39/DeepNet/issues/39"}