Clojure IllegalArgumentException:defprotocol中没有单个方法

我使用

defprotocol

实现多态性。我有一个简单的逻辑门。

forward

和

backward

是每个闸门要实现的两个功能

代码：

;; a single unit has forward value and backward gradient.
(defrecord Unit 
    [value gradient])

(defrecord Gate
    [^:Unit input-a ^:Unit input-b])

(defprotocol GateOps
  (forward [this])
  (backward [this back-grad]))

(extend-protocol GateOps
  Unit
  (forward [this]
    (-> this :value))
  (backward [this back-grad]
    ({this back-grad})))

(defrecord MultiplyGate [input-a input-b]
  GateOps
  (forward [this]
    (* (forward (-> this :input-a)) (forward (:input-b this))))

  (backward [this back-grad]
    (let [val-a (forward (-> this :input-a))
          val-b (forward (-> this :input-b))
          input-a (:input-a this)
          input-b (:input-b this)]
      (merge-with + (backward input-a (* val-b back-grad))
                    (backward input-b (* val-a back-grad))))))

(defrecord AddGate [input-a input-b]
  GateOps
  (forward [this]
    (+ (forward (:input-a this)) (forward (:input-b this))))

  (backward [this back-grad]
    (let [val-a (forward (-> this :input-a))
          val-b (forward (-> this :input-b))
          input-a (:input-a this)
          input-b (:input-b this)]
      (merge-with + (backward input-a (* 1.0 back-grad))
                    (backward input-b (* 1.0 back-grad))))))


(defn sig [x]
  (/ 1 (+ 1 (Math/pow Math/E (- x)))))

(defrecord SigmoidGate [gate]
  GateOps
  (forward [this]
    (sig (forward (:gate this))))

  (backward [this back-grad]
    (let [s (forward this)
          ds (* s (- 1 s))]
      (backward (:gate this) ds))))

当

forward

工作正常时，在

backward

中，我遇到异常：

user> (neurals.core/forward neurals.core/sigaxcby)
0.8807970779778823

user> (neurals.core/backward neurals.core/sigaxcby)
CompilerException java.lang.IllegalArgumentException: 
No single method: backward of interface: neurals.core.GateOps 
found for function: backward of protocol: GateOps, 
 compiling:(C:\xxyyzz\AppData\Local\Temp\form-init8132866244624247216.clj:1:1) 

user> 

版本信息：Clojure 1.6.0

---

我做错了什么？

这很奇怪，因为

错误输出

与实际错误没有任何关联：

算术错误

我应该为

向后

函数提供初始

向后梯度

值：

(neurals.core/backward neurals.core/sigaxcby 1.0) 

---

知道了。。也应该给落后以最初的价值。少了一个参数。是的，这很奇怪，但“没有单一方法”通常意味着方法上的算术错误。。谢谢会将此错误添加到我的错误命名空间中。：）