diff --git a/Cargo.toml b/Cargo.toml
index 60ca258..7a8234d 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -13,9 +13,20 @@ keywords = ["finite", "float", "hash", "nan", "ord"]
 travis-ci = { repository = "olson-sean-k/decorum" }
 
 [features]
-default = ["serialize-serde", "std"]
+default = ["approx", "serialize-serde", "std"]
 serialize-serde = ["serde", "serde_derive"]
-std = ["num-traits/std"]
+std = ["approx/std", "num-traits/std"]
+
+[dependencies.approx]
+version = "^0.3.0"
+default-features = false
+features = []
+optional = true
+
+[dependencies.num-traits]
+version = "^0.2.4"
+default-features = false
+features = []
 
 [dependencies.serde]
 version = "1.0"
@@ -27,10 +38,5 @@ version = "1.0"
 default-features = false
 optional = true
 
-[dependencies.num-traits]
-version = "^0.2.4"
-default-features = false
-features = []
-
 [dev-dependencies]
 num = "^0.2.0"
diff --git a/src/proxy.rs b/src/proxy.rs
index a31ffc2..d9d0f6e 100644
--- a/src/proxy.rs
+++ b/src/proxy.rs
@@ -1,3 +1,4 @@
+use approx::{AbsDiffEq, RelativeEq, UlpsEq};
 use core::cmp::Ordering;
 use core::fmt::{self, Display, Formatter, LowerExp, UpperExp};
 use core::hash::{Hash, Hasher};
@@ -212,7 +213,7 @@ where
 
 // It is not possible to implement `From` for proxies in a generic way, because
 // the `FloatConstraint` types `T` and `U` may be the same and conflict with
-// the reflexive implementation in core. A similar problem prevents
+// the reflexive implementation in `core`. A similar problem prevents
 // implementing `From` over a type `T: Float`.
 
 impl<T> From<NotNan<T>> for Ordered<T>
@@ -270,6 +271,23 @@ where
     }
 }
 
+impl<T, P> AbsDiffEq for ConstrainedFloat<T, P>
+where
+    T: AbsDiffEq<Epsilon = T> + Float + Primitive,
+    P: FloatConstraint<T> + ConstraintEq<T>,
+{
+    type Epsilon = Self;
+
+    fn default_epsilon() -> Self::Epsilon {
+        Self::from_inner(T::default_epsilon())
+    }
+
+    fn abs_diff_eq(&self, other: &Self, epsilon: Self::Epsilon) -> bool {
+        self.into_inner()
+            .abs_diff_eq(&other.into_inner(), epsilon.into_inner())
+    }
+}
+
 impl<T, P> Add for ConstrainedFloat<T, P>
 where
     T: Float + Primitive,
@@ -1273,6 +1291,29 @@ where
     }
 }
 
+impl<T, P> RelativeEq for ConstrainedFloat<T, P>
+where
+    T: Float + Primitive + RelativeEq<Epsilon = T>,
+    P: FloatConstraint<T> + ConstraintEq<T>,
+{
+    fn default_max_relative() -> Self::Epsilon {
+        Self::from_inner(T::default_max_relative())
+    }
+
+    fn relative_eq(
+        &self,
+        other: &Self,
+        epsilon: Self::Epsilon,
+        max_relative: Self::Epsilon,
+    ) -> bool {
+        self.into_inner().relative_eq(
+            &other.into_inner(),
+            epsilon.into_inner(),
+            max_relative.into_inner(),
+        )
+    }
+}
+
 impl<T, P> Rem for ConstrainedFloat<T, P>
 where
     T: Float + Primitive,
@@ -1468,6 +1509,21 @@ where
     }
 }
 
+impl<T, P> UlpsEq for ConstrainedFloat<T, P>
+where
+    T: Float + Primitive + UlpsEq<Epsilon = T>,
+    P: FloatConstraint<T> + ConstraintEq<T>,
+{
+    fn default_max_ulps() -> u32 {
+        T::default_max_ulps()
+    }
+
+    fn ulps_eq(&self, other: &Self, epsilon: Self::Epsilon, max_ulps: u32) -> bool {
+        self.into_inner()
+            .ulps_eq(&other.into_inner(), epsilon.into_inner(), max_ulps)
+    }
+}
+
 impl<T, P> UpperExp for ConstrainedFloat<T, P>
 where
     T: Float + UpperExp + Primitive,